Point compose to remote Elasticsearch and Qdrant

- Add clickable badges for backlinks and references that trigger query string searches
- Improve toggle checkbox layout with better styling
- Add description block styling with scrollable container
- Update results styling with bordered cards and shadows
- Add favicon support across pages
- Enhance .env loading with logging for debugging

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

.dockerignore

@@ -0,0 +1,11 @@
+.git
+.gitignore
+.venv
+__pycache__
+*.pyc
+*.pyo
+.DS_Store
+node_modules
+data
+videos
+*.log

AGENTS.md

@@ -0,0 +1,31 @@
+# Repository Guidelines
+
+## Project Structure & Module Organization
+- Core modules live under `python_app/`: `config.py` centralizes settings, `transcript_collector.py` gathers transcripts, `ingest.py` handles Elasticsearch bulk loads, and `search_app.py` exposes the Flask UI.
+- Static assets belong in `static/` (`index.html`, `frequency.html`, companion JS/CSS). Keep HTML here and wire it up through Flask routes.
+- Runtime artifacts land in `data/` (`raw/` for downloads, `video_metadata/` for cleaned payloads). Preserve the JSON schema emitted by the collector.
+- When adding utilities, place them in `python_app/` and use package-relative imports so scripts continue to run via `python -m`.
+
+## Build, Test, and Development Commands
+- `python -m venv .venv && source .venv/bin/activate`: bootstrap the virtualenv used by all scripts.
+- `pip install -r requirements.txt`: install Flask, Elasticsearch tooling, Google API clients, and dotenv support.
+- `python -m python_app.transcript_collector --channel UC... --output data/raw`: fetch transcript JSON for a channel; rerun to refresh cached data.
+- `python -m python_app.ingest --source data/video_metadata --index this_little_corner_py`: index prepared metadata and auto-create mappings when needed.
+- `python -m python_app.search_app`: launch the Flask server on port 8080 for UI smoke tests.
+
+## Coding Style & Naming Conventions
+- Follow PEP 8 with 4-space indentation, `snake_case` for functions/modules, and `CamelCase` for classes; reserve UPPER_SNAKE_CASE for configuration constants.
+- Keep Elasticsearch payload keys lower-case with underscores, and centralize shared values in `config.py` rather than scattering literals.
+
+## Testing Guidelines
+- No automated suite is committed yet; when adding coverage, create `tests/` modules using `pytest` with files named `test_*.py`.
+- Focus tests on collector pagination, ingest transformations, and Flask route helpers, and run `python -m pytest` locally before opening a PR.
+- Manually verify by ingesting a small sample into a local Elasticsearch node and checking facets, highlights, and transcript retrieval via the UI.
+
+## Commit & Pull Request Guidelines
+- Mirror the existing history: short, imperative commit subjects (e.g. “Fix results overflow”, “Add video reference tracking”).
+- PRs should describe scope, list environment variables or indices touched, link issues, and attach before/after screenshots whenever UI output changes. Highlight Elasticsearch mapping or data migration impacts for both search and frontend reviewers.
+
+## Configuration & Security Tips
+- Load credentials through environment variables (`ELASTIC_URL`, `ELASTIC_USERNAME`, `ELASTIC_PASSWORD`, `ELASTIC_API_KEY`, `YOUTUBE_API_KEY`) or a `.env` file, and keep secrets out of version control.
+- Adjust `ELASTIC_VERIFY_CERTS`, `ELASTIC_CA_CERT`, and `ELASTIC_DEBUG` only while debugging, and prefer branch-specific indices (`this_little_corner_py_<initials>`) to avoid clobbering shared data.

Dockerfile

@@ -0,0 +1,32 @@
+FROM python:3.11-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1
+ENV PYTHONUNBUFFERED=1
+
+WORKDIR /app
+
+# System deps kept lean to support torch/sentence-transformers wheels.
+RUN apt-get update \
+  && apt-get install -y --no-install-recommends build-essential git curl \
+  && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -r /app/requirements.txt
+
+# Copy the package into /app/python_app so `python -m python_app.search_app` works.
+COPY . /app/python_app
+
+ENV ELASTIC_URL=http://elasticsearch:9200 \
+    ELASTIC_INDEX=this_little_corner_py \
+    ELASTIC_VERIFY_CERTS=0 \
+    QDRANT_URL=http://qdrant:6333 \
+    QDRANT_COLLECTION=tlc-captions-full \
+    QDRANT_VECTOR_NAME= \
+    QDRANT_VECTOR_SIZE=1024 \
+    QDRANT_EMBED_MODEL=BAAI/bge-large-en-v1.5 \
+    LOCAL_DATA_DIR=/app/data/video_metadata
+
+EXPOSE 8080
+
+WORKDIR /app
+CMD ["python", "-m", "python_app.search_app"]

README.md

@@ -85,3 +85,22 @@ Visit <http://localhost:8080/> and you’ll see a barebones UI that:
 
 Feel free to expand on this scaffold—add proper logging, schedule transcript
 updates, or flesh out the UI—once you’re happy with the baseline behaviour.
+
+## Run with Docker Compose (App Only; Remote ES/Qdrant)
+
+The provided compose file builds/runs only the Flask app and expects **remote** Elasticsearch/Qdrant endpoints. Supply them via environment variables (directly or a `.env` alongside `docker-compose.yml`):
+
+```bash
+ELASTIC_URL=https://your-es-host:9200 \
+QDRANT_URL=https://your-qdrant-host:6333 \
+docker compose up --build
+```
+
+Other tunables (defaults shown in compose):
+- `ELASTIC_INDEX` (default `this_little_corner_py`)
+- `ELASTIC_USERNAME` / `ELASTIC_PASSWORD` or `ELASTIC_API_KEY`
+- `ELASTIC_VERIFY_CERTS` (set to `1` for real TLS verification)
+- `QDRANT_COLLECTION` (default `tlc-captions-full`)
+- `QDRANT_VECTOR_NAME` / `QDRANT_VECTOR_SIZE` / `QDRANT_EMBED_MODEL`
+
+Port 8080 on the host is forwarded to the app. Mount `./data` (read-only) if you want local fallbacks for metrics (`LOCAL_DATA_DIR=/app/data/video_metadata`); otherwise the app will rely purely on the remote backends. Stop the container with `docker compose down`.

config.py

@@ -16,6 +16,20 @@ from dataclasses import dataclass
 from pathlib import Path
 from typing import Optional
 
+# Load .env file if it exists
+try:
+    from dotenv import load_dotenv
+    import logging
+
+    _logger = logging.getLogger(__name__)
+    _env_path = Path(__file__).parent / ".env"
+    if _env_path.exists():
+        _logger.info("Loading .env from: %s", _env_path)
+        result = load_dotenv(_env_path, override=True)
+        _logger.info("load_dotenv result: %s", result)
+except ImportError:
+    pass  # python-dotenv not installed
+
 
 @dataclass(frozen=True)
 class ElasticSettings:
@@ -44,6 +58,11 @@ class AppConfig:
     elastic: ElasticSettings
     data: DataSettings
     youtube: YoutubeSettings
+    qdrant_url: str
+    qdrant_collection: str
+    qdrant_vector_name: Optional[str]
+    qdrant_vector_size: int
+    qdrant_embed_model: str
 
 
 def _env(name: str, default: Optional[str] = None) -> Optional[str]:
@@ -75,7 +94,16 @@ def load_config() -> AppConfig:
     )
     data = DataSettings(root=data_root)
     youtube = YoutubeSettings(api_key=_env("YOUTUBE_API_KEY"))
-    return AppConfig(elastic=elastic, data=data, youtube=youtube)
+    return AppConfig(
+        elastic=elastic,
+        data=data,
+        youtube=youtube,
+        qdrant_url=_env("QDRANT_URL", "http://localhost:6333"),
+        qdrant_collection=_env("QDRANT_COLLECTION", "tlc_embeddings"),
+        qdrant_vector_name=_env("QDRANT_VECTOR_NAME"),
+        qdrant_vector_size=int(_env("QDRANT_VECTOR_SIZE", "1024")),
+        qdrant_embed_model=_env("QDRANT_EMBED_MODEL", "BAAI/bge-large-en-v1.5"),
+    )
 
 
 CONFIG = load_config()

docker-compose.yml

@@ -0,0 +1,26 @@
+version: "3.9"
+
+# Runs only the Flask app container, pointing to remote Elasticsearch/Qdrant.
+# Provide ELASTIC_URL / QDRANT_URL (and related) via environment or a .env file.
+services:
+  app:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    ports:
+      - "8080:8080"
+    environment:
+      ELASTIC_URL: ${ELASTIC_URL:?set ELASTIC_URL to your remote Elasticsearch URL}
+      ELASTIC_INDEX: ${ELASTIC_INDEX:-this_little_corner_py}
+      ELASTIC_USERNAME: ${ELASTIC_USERNAME:-}
+      ELASTIC_PASSWORD: ${ELASTIC_PASSWORD:-}
+      ELASTIC_API_KEY: ${ELASTIC_API_KEY:-}
+      ELASTIC_VERIFY_CERTS: ${ELASTIC_VERIFY_CERTS:-0}
+      QDRANT_URL: ${QDRANT_URL:?set QDRANT_URL to your remote Qdrant URL}
+      QDRANT_COLLECTION: ${QDRANT_COLLECTION:-tlc-captions-full}
+      QDRANT_VECTOR_NAME: ${QDRANT_VECTOR_NAME:-}
+      QDRANT_VECTOR_SIZE: ${QDRANT_VECTOR_SIZE:-1024}
+      QDRANT_EMBED_MODEL: ${QDRANT_EMBED_MODEL:-BAAI/bge-large-en-v1.5}
+      LOCAL_DATA_DIR: ${LOCAL_DATA_DIR:-/app/data/video_metadata}
+    volumes:
+      - ./data:/app/data:ro

ingest.py

@@ -90,6 +90,10 @@ def build_bulk_actions(
                 "transcript_full": transcript_full,
                 "transcript_secondary_full": doc.get("transcript_secondary_full"),
                 "transcript_parts": parts,
+                "internal_references": doc.get("internal_references", []),
+                "internal_references_count": doc.get("internal_references_count", 0),
+                "referenced_by": doc.get("referenced_by", []),
+                "referenced_by_count": doc.get("referenced_by_count", 0),
             },
         }
 
@@ -121,6 +125,10 @@ def ensure_index(client: "Elasticsearch", index: str) -> None:
                         "text": {"type": "text"},
                     },
                 },
+                "internal_references": {"type": "keyword"},
+                "internal_references_count": {"type": "integer"},
+                "referenced_by": {"type": "keyword"},
+                "referenced_by_count": {"type": "integer"},
             }
         },
     )

requirements.txt

@@ -2,3 +2,6 @@ Flask>=2.3
 elasticsearch>=7.0.0,<9.0.0
 youtube-transcript-api>=0.6
 google-api-python-client>=2.0.0
+python-dotenv>=0.19.0
+requests>=2.31.0
+sentence-transformers>=2.7.0

search_app.py

@@ -1,11 +1,15 @@
 """
-Flask application exposing a minimal search API backed by Elasticsearch.
+Flask application exposing search, graph, and transcript endpoints for TLC.
 
 Routes:
-    GET /                -> Static HTML search page.
+    GET /                  -> static HTML search page.
-    GET /api/channels    -> List available channels (via terms aggregation).
+    GET /graph             -> static reference graph UI.
-    GET /api/search      -> Search index with pagination and simple highlighting.
+    GET /vector-search     -> experimental Qdrant vector search UI.
-    GET /api/transcript  -> Return full transcript for a given video_id.
+    GET /api/channels      -> channels aggregation.
+    GET /api/search        -> Elasticsearch keyword search.
+    POST /api/vector-search -> Qdrant vector similarity query.
+    GET /api/graph         -> reference graph API.
+    GET /api/transcript    -> transcript JSON payload.
 """
 
 from __future__ import annotations
@@ -15,13 +19,20 @@ import json
 import logging
 import re
 from pathlib import Path
-from typing import Any, Dict, Iterable, List, Optional, Sequence, Set
+from typing import Any, Dict, Iterable, List, Optional, Sequence, Set, Tuple
 
-from collections import Counter
+from collections import Counter, deque
 from datetime import datetime
 
 from flask import Flask, jsonify, request, send_from_directory
 
+import requests
+
+try:
+    from sentence_transformers import SentenceTransformer  # type: ignore
+except ImportError:  # pragma: no cover - optional dependency
+    SentenceTransformer = None
+
 from .config import CONFIG, AppConfig
 
 try:
@@ -32,6 +43,35 @@ except ImportError:  # pragma: no cover - dependency optional
     BadRequestError = Exception  # type: ignore
 
 LOGGER = logging.getLogger(__name__)
+_EMBED_MODEL = None
+_EMBED_MODEL_NAME: Optional[str] = None
+
+
+def _ensure_embedder(model_name: str) -> "SentenceTransformer":
+    global _EMBED_MODEL, _EMBED_MODEL_NAME
+    if SentenceTransformer is None:  # pragma: no cover - optional dependency
+        raise RuntimeError(
+            "sentence-transformers is required for vector search. Install via pip install sentence-transformers."
+        )
+    if _EMBED_MODEL is None or _EMBED_MODEL_NAME != model_name:
+        LOGGER.info("Loading embedding model: %s", model_name)
+        _EMBED_MODEL = SentenceTransformer(model_name)
+        _EMBED_MODEL_NAME = model_name
+    return _EMBED_MODEL
+
+
+def embed_query(text: str, *, model_name: str, expected_dim: int) -> List[float]:
+    embedder = _ensure_embedder(model_name)
+    vector = embedder.encode(
+        [f"query: {text}"],
+        show_progress_bar=False,
+        normalize_embeddings=True,
+    )[0].tolist()
+    if len(vector) != expected_dim:
+        raise RuntimeError(
+            f"Embedding dimension mismatch (expected {expected_dim}, got {len(vector)})"
+        )
+    return vector
 
 
 def _ensure_client(config: AppConfig) -> "Elasticsearch":
@@ -286,6 +326,24 @@ def parse_channel_params(values: Iterable[Optional[str]]) -> List[str]:
     return channels
 
 
+def build_year_filter(year: Optional[str]) -> Optional[Dict]:
+    if not year:
+        return None
+    try:
+        year_int = int(year)
+        return {
+            "range": {
+                "date": {
+                    "gte": f"{year_int}-01-01",
+                    "lt": f"{year_int + 1}-01-01",
+                    "format": "yyyy-MM-dd"
+                }
+            }
+        }
+    except (ValueError, TypeError):
+        return None
+
+
 def build_channel_filter(channels: Optional[Sequence[str]]) -> Optional[Dict]:
     if not channels:
         return None
@@ -320,6 +378,7 @@ def build_query_payload(
     query: str,
     *,
     channels: Optional[Sequence[str]] = None,
+    year: Optional[str] = None,
     sort: str = "relevant",
     use_exact: bool = True,
     use_fuzzy: bool = True,
@@ -333,6 +392,10 @@ def build_query_payload(
     if channel_filter:
         filters.append(channel_filter)
 
+    year_filter = build_year_filter(year)
+    if year_filter:
+        filters.append(year_filter)
+
     if use_query_string:
         base_fields = ["title^3", "description^2", "transcript_full", "transcript_secondary_full"]
         qs_query = (query or "").strip() or "*"
@@ -376,6 +439,8 @@ def build_query_payload(
             body["sort"] = [{"date": {"order": "desc"}}]
         elif sort == "older":
             body["sort"] = [{"date": {"order": "asc"}}]
+        elif sort == "referenced":
+            body["sort"] = [{"referenced_by_count": {"order": "desc", "unmapped_type": "long"}}]
         return body
 
     if query:
@@ -403,6 +468,17 @@ def build_query_payload(
                     }
                 }
             )
+            should.append(
+                {
+                    "match_phrase": {
+                        "title": {
+                            "query": query,
+                            "slop": 0,
+                            "boost": 50.0,
+                        }
+                    }
+                }
+            )
         if use_fuzzy:
             should.append(
                 {
@@ -479,6 +555,8 @@ def build_query_payload(
         body["sort"] = [{"date": {"order": "desc"}}]
     elif sort == "older":
         body["sort"] = [{"date": {"order": "asc"}}]
+    elif sort == "referenced":
+        body["sort"] = [{"referenced_by_count": {"order": "desc", "unmapped_type": "long"}}]
     return body
 
 
@@ -486,15 +564,182 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
     app = Flask(__name__, static_folder=str(Path(__file__).parent / "static"))
     client = _ensure_client(config)
     index = config.elastic.index
+    qdrant_url = config.qdrant_url
+    qdrant_collection = config.qdrant_collection
+    qdrant_vector_name = config.qdrant_vector_name
+    qdrant_vector_size = config.qdrant_vector_size
+    qdrant_embed_model = config.qdrant_embed_model
 
     @app.route("/")
     def index_page():
         return send_from_directory(app.static_folder, "index.html")
 
+    @app.route("/graph")
+    def graph_page():
+        return send_from_directory(app.static_folder, "graph.html")
+
+    @app.route("/vector-search")
+    def vector_search_page():
+        return send_from_directory(app.static_folder, "vector.html")
+
     @app.route("/static/<path:filename>")
     def static_files(filename: str):
         return send_from_directory(app.static_folder, filename)
 
+    def normalize_reference_list(values: Any) -> List[str]:
+        if values is None:
+            return []
+        if isinstance(values, (list, tuple, set)):
+            iterable = values
+        else:
+            iterable = [values]
+        normalized: List[str] = []
+        for item in iterable:
+            candidate: Optional[str]
+            if isinstance(item, dict):
+                candidate = item.get("video_id") or item.get("id")  # type: ignore[assignment]
+            else:
+                candidate = item  # type: ignore[assignment]
+            if candidate is None:
+                continue
+            text = str(candidate).strip()
+            if not text:
+                continue
+            if text.lower() in {"none", "null"}:
+                continue
+            normalized.append(text)
+        return normalized
+
+    def build_graph_payload(
+        root_id: str, depth: int, max_nodes: int
+    ) -> Dict[str, Any]:
+        root_id = root_id.strip()
+        if not root_id:
+            return {"nodes": [], "links": [], "root": root_id, "depth": depth, "meta": {}}
+
+        doc_cache: Dict[str, Optional[Dict[str, Any]]] = {}
+
+        def fetch_document(video_id: str) -> Optional[Dict[str, Any]]:
+            if video_id in doc_cache:
+                return doc_cache[video_id]
+            try:
+                result = client.get(index=index, id=video_id)
+                doc_cache[video_id] = result.get("_source")
+            except Exception as exc:  # pragma: no cover - elasticsearch handles errors
+                LOGGER.debug("Graph: failed to load %s: %s", video_id, exc)
+                doc_cache[video_id] = None
+            return doc_cache[video_id]
+
+        nodes: Dict[str, Dict[str, Any]] = {}
+        links: List[Dict[str, Any]] = []
+        link_seen: Set[Tuple[str, str, str]] = set()
+        queue: deque[Tuple[str, int]] = deque([(root_id, 0)])
+        queued: Set[str] = {root_id}
+        visited: Set[str] = set()
+
+        while queue and len(nodes) < max_nodes:
+            current_id, level = queue.popleft()
+            queued.discard(current_id)
+            if current_id in visited:
+                continue
+            doc = fetch_document(current_id)
+            if doc is None:
+                if current_id == root_id:
+                    break
+                visited.add(current_id)
+                continue
+
+            visited.add(current_id)
+            nodes[current_id] = {
+                "id": current_id,
+                "title": doc.get("title") or current_id,
+                "channel_id": doc.get("channel_id"),
+                "channel_name": doc.get("channel_name") or doc.get("channel_id") or "Unknown",
+                "url": doc.get("url"),
+                "date": doc.get("date"),
+                "is_root": current_id == root_id,
+            }
+
+            if level >= depth:
+                continue
+
+            neighbor_ids: List[str] = []
+
+            for ref_id in normalize_reference_list(doc.get("internal_references")):
+                if ref_id == current_id:
+                    continue
+                key = (current_id, ref_id, "references")
+                if key not in link_seen:
+                    links.append(
+                        {"source": current_id, "target": ref_id, "relation": "references"}
+                    )
+                    link_seen.add(key)
+                neighbor_ids.append(ref_id)
+
+            for ref_id in normalize_reference_list(doc.get("referenced_by")):
+                if ref_id == current_id:
+                    continue
+                key = (ref_id, current_id, "referenced_by")
+                if key not in link_seen:
+                    links.append(
+                        {"source": ref_id, "target": current_id, "relation": "referenced_by"}
+                    )
+                    link_seen.add(key)
+                neighbor_ids.append(ref_id)
+
+            for neighbor in neighbor_ids:
+                if neighbor in visited or neighbor in queued:
+                    continue
+                if len(nodes) + len(queue) >= max_nodes:
+                    break
+                queue.append((neighbor, level + 1))
+                queued.add(neighbor)
+
+        # Ensure nodes referenced by links exist in the payload.
+        for link in links:
+            for key in ("source", "target"):
+                node_id = link[key]
+                if node_id in nodes:
+                    continue
+                doc = fetch_document(node_id)
+                if doc is None:
+                    nodes[node_id] = {
+                        "id": node_id,
+                        "title": node_id,
+                        "channel_id": None,
+                        "channel_name": "Unknown",
+                        "url": None,
+                        "date": None,
+                        "is_root": node_id == root_id,
+                    }
+                else:
+                    nodes[node_id] = {
+                        "id": node_id,
+                        "title": doc.get("title") or node_id,
+                        "channel_id": doc.get("channel_id"),
+                        "channel_name": doc.get("channel_name") or doc.get("channel_id") or "Unknown",
+                        "url": doc.get("url"),
+                        "date": doc.get("date"),
+                        "is_root": node_id == root_id,
+                    }
+
+        links = [
+            link
+            for link in links
+            if link.get("source") in nodes and link.get("target") in nodes
+        ]
+
+        return {
+            "root": root_id,
+            "depth": depth,
+            "nodes": list(nodes.values()),
+            "links": links,
+            "meta": {
+                "node_count": len(nodes),
+                "link_count": len(links),
+            },
+        }
+
     @app.route("/api/channels")
     def channels():
         base_channels_body = {
@@ -553,21 +798,99 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
             .get("channels", {})
             .get("buckets", [])
         )
+        data = []
+        for bucket in buckets:
+            key = bucket.get("key")
+            name_hit = (
+                bucket.get("name", {})
+                .get("hits", {})
+                .get("hits", [{}])[0]
+                .get("_source", {})
+                .get("channel_name")
+            )
+            display_name = name_hit or key or "Unknown"
+            data.append(
+                {
+                    "Id": key,
+                    "Name": display_name,
+                    "Count": bucket.get("doc_count", 0),
+                }
+            )
+        data.sort(key=lambda item: item["Name"].lower())
+        return jsonify(data)
+
+    @app.route("/api/graph")
+    def graph_api():
+        video_id = (request.args.get("video_id") or "").strip()
+        if not video_id:
+            return jsonify({"error": "video_id is required"}), 400
+
+        try:
+            depth = int(request.args.get("depth", "1"))
+        except ValueError:
+            depth = 1
+        depth = max(0, min(depth, 3))
+
+        try:
+            max_nodes = int(request.args.get("max_nodes", "200"))
+        except ValueError:
+            max_nodes = 200
+        max_nodes = max(10, min(max_nodes, 400))
+
+        payload = build_graph_payload(video_id, depth, max_nodes)
+        if not payload["nodes"]:
+            return (
+                jsonify({"error": f"Video '{video_id}' was not found in the index."}),
+                404,
+            )
+        payload["meta"]["max_nodes"] = max_nodes
+        return jsonify(payload)
+
+    @app.route("/api/years")
+    def years():
+        body = {
+            "size": 0,
+            "aggs": {
+                "years": {
+                    "date_histogram": {
+                        "field": "date",
+                        "calendar_interval": "year",
+                        "format": "yyyy",
+                        "order": {"_key": "desc"}
+                    }
+                }
+            }
+        }
+
+        if config.elastic.debug:
+            LOGGER.info(
+                "Elasticsearch years request: %s",
+                json.dumps({"index": index, "body": body}, indent=2),
+            )
+
+        response = client.search(index=index, body=body)
+
+        if config.elastic.debug:
+            LOGGER.info(
+                "Elasticsearch years response: %s",
+                json.dumps(response, indent=2, default=str),
+            )
+
+        buckets = (
+            response.get("aggregations", {})
+            .get("years", {})
+            .get("buckets", [])
+        )
+
         data = [
             {
-                "Id": bucket.get("key"),
+                "Year": bucket.get("key_as_string"),
-                "Name": (
-                    bucket.get("name", {})
-                    .get("hits", {})
-                    .get("hits", [{}])[0]
-                    .get("_source", {})
-                    .get("channel_name", bucket.get("key"))
-                ),
                 "Count": bucket.get("doc_count", 0),
             }
             for bucket in buckets
+            if bucket.get("doc_count", 0) > 0
         ]
-        data.sort(key=lambda item: item["Name"].lower())
+
         return jsonify(data)
 
     @app.route("/api/search")
@@ -578,6 +901,7 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
         if legacy_channel:
             raw_channels.append(legacy_channel)
         channels = parse_channel_params(raw_channels)
+        year = request.args.get("year", "", type=str) or None
         sort = request.args.get("sort", "relevant", type=str)
         page = max(request.args.get("page", 0, type=int), 0)
         size = max(request.args.get("size", 10, type=int), 1)
@@ -598,6 +922,7 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
         payload = build_query_payload(
             query,
             channels=channels,
+            year=year,
             sort=sort,
             use_exact=use_exact,
             use_fuzzy=use_fuzzy,
@@ -642,10 +967,13 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
         for hit in hits.get("hits", []):
             source = hit.get("_source", {})
             highlight_map = hit.get("highlight", {})
-            transcript_highlight = (
+            transcript_highlight = [
-                (highlight_map.get("transcript_full", []) or [])
+                {"html": value, "source": "primary"}
-                + (highlight_map.get("transcript_secondary_full", []) or [])
+                for value in (highlight_map.get("transcript_full", []) or [])
-            )
+            ] + [
+                {"html": value, "source": "secondary"}
+                for value in (highlight_map.get("transcript_secondary_full", []) or [])
+            ]
 
             title_html = (
                 highlight_map.get("title")
@@ -665,12 +993,18 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
                     "description": source.get("description"),
                     "descriptionHtml": description_html,
                     "date": source.get("date"),
+                    "duration": source.get("duration"),
                     "url": source.get("url"),
                     "toHighlight": transcript_highlight,
                     "highlightSource": {
                         "primary": bool(highlight_map.get("transcript_full")),
                         "secondary": bool(highlight_map.get("transcript_secondary_full")),
                     },
+                    "internal_references_count": source.get("internal_references_count", 0),
+                    "internal_references": source.get("internal_references", []),
+                    "referenced_by_count": source.get("referenced_by_count", 0),
+                    "referenced_by": source.get("referenced_by", []),
+                    "video_status": source.get("video_status"),
                 }
             )
 
@@ -716,6 +1050,7 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
         if legacy_channel:
             raw_channels.append(legacy_channel)
         channels = parse_channel_params(raw_channels)
+        year = request.args.get("year", "", type=str) or None
         interval = (request.args.get("interval", "month") or "month").lower()
         allowed_intervals = {"day", "week", "month", "quarter", "year"}
         if interval not in allowed_intervals:
@@ -723,45 +1058,50 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
         start = request.args.get("start", type=str)
         end = request.args.get("end", type=str)
 
-        filters: List[Dict] = []
+        def parse_flag(name: str, default: bool = True) -> bool:
-        channel_filter = build_channel_filter(channels)
+            value = request.args.get(name)
-        if channel_filter:
+            if value is None:
-            filters.append(channel_filter)
+                return default
+            lowered = value.lower()
+            return lowered not in {"0", "false", "no"}
+
+        use_exact = parse_flag("exact", True)
+        use_fuzzy = parse_flag("fuzzy", True)
+        use_phrase = parse_flag("phrase", True)
+        if use_query_string:
+            use_exact = use_fuzzy = use_phrase = False
+
+        search_payload = build_query_payload(
+            term,
+            channels=channels,
+            year=year,
+            sort="relevant",
+            use_exact=use_exact,
+            use_fuzzy=use_fuzzy,
+            use_phrase=use_phrase,
+            use_query_string=use_query_string,
+        )
+        query = search_payload.get("query", {"match_all": {}})
+
         if start or end:
             range_filter: Dict[str, Dict[str, Dict[str, str]]] = {"range": {"date": {}}}
             if start:
                 range_filter["range"]["date"]["gte"] = start
             if end:
                 range_filter["range"]["date"]["lte"] = end
-            filters.append(range_filter)
+            if "bool" in query:
-
+                bool_clause = query.setdefault("bool", {})
-        base_fields = ["title^3", "description^2", "transcript_full", "transcript_secondary_full"]
+                existing_filter = bool_clause.get("filter")
-        if use_query_string:
+                if existing_filter is None:
-            qs_query = term or "*"
+                    bool_clause["filter"] = [range_filter]
-            must_clause: List[Dict[str, Any]] = [
+                elif isinstance(existing_filter, list):
-                {
+                    bool_clause["filter"].append(range_filter)
-                    "query_string": {
+                else:
-                        "query": qs_query,
+                    bool_clause["filter"] = [existing_filter, range_filter]
-                        "default_operator": "AND",
+            elif query.get("match_all") is not None:
-                        "fields": base_fields,
+                query = {"bool": {"filter": [range_filter]}}
-                    }
+            else:
-                }
+                query = {"bool": {"must": [query], "filter": [range_filter]}}
-            ]
-        else:
-            must_clause = [
-                {
-                    "multi_match": {
-                        "query": term,
-                        "fields": base_fields,
-                        "type": "best_fields",
-                        "operator": "and",
-                    }
-                }
-            ]
-
-        query: Dict[str, Any] = {"bool": {"must": must_clause}}
-        if filters:
-            query["bool"]["filter"] = filters
 
         histogram: Dict[str, Any] = {
             "field": "date",
@@ -791,7 +1131,15 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
                                 "field": "channel_id.keyword",
                                 "size": channel_terms_size,
                                 "order": {"_count": "desc"},
-                            }
+                            },
+                            "aggs": {
+                                "channel_name_hit": {
+                                    "top_hits": {
+                                        "size": 1,
+                                        "_source": {"includes": ["channel_name"]},
+                                    }
+                                }
+                            },
                         }
                     },
                 }
@@ -830,7 +1178,7 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
             .get("buckets", [])
         )
 
-        channel_totals: Dict[str, int] = {}
+        channel_totals: Dict[str, Dict[str, Any]] = {}
         buckets: List[Dict[str, Any]] = []
         for bucket in raw_buckets:
             date_str = bucket.get("key_as_string")
@@ -840,14 +1188,28 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
                 cid = ch_bucket.get("key")
                 count = ch_bucket.get("doc_count", 0)
                 if cid:
-                    channel_entries.append({"id": cid, "count": count})
+                    hit_source = (
-                    channel_totals[cid] = channel_totals.get(cid, 0) + count
+                        ch_bucket.get("channel_name_hit", {})
+                        .get("hits", {})
+                        .get("hits", [{}])[0]
+                        .get("_source", {})
+                    )
+                    channel_name = hit_source.get("channel_name") if isinstance(hit_source, dict) else None
+                    channel_entries.append({"id": cid, "count": count, "name": channel_name})
+                    if cid not in channel_totals:
+                        channel_totals[cid] = {"total": 0, "name": channel_name}
+                    channel_totals[cid]["total"] += count
+                    if channel_name and not channel_totals[cid].get("name"):
+                        channel_totals[cid]["name"] = channel_name
             buckets.append(
                 {"date": date_str, "total": total, "channels": channel_entries}
             )
 
         ranked_channels = sorted(
-            [{"id": cid, "total": total} for cid, total in channel_totals.items()],
+            [
+                {"id": cid, "total": info.get("total", 0), "name": info.get("name")}
+                for cid, info in channel_totals.items()
+            ],
             key=lambda item: item["total"],
             reverse=True,
         )
@@ -867,6 +1229,145 @@ def create_app(config: AppConfig = CONFIG) -> Flask:
     def frequency_page():
         return send_from_directory(app.static_folder, "frequency.html")
 
+    @app.route("/api/vector-search", methods=["POST"])
+    def api_vector_search():
+        payload = request.get_json(silent=True) or {}
+        query_text = (payload.get("query") or "").strip()
+        filters = payload.get("filters") or {}
+        limit = max(int(payload.get("size", 10)), 1)
+        offset = max(int(payload.get("offset", 0)), 0)
+
+        if not query_text:
+            return jsonify(
+                {"items": [], "totalResults": 0, "offset": offset, "error": "empty_query"}
+            )
+
+        try:
+            query_vector = embed_query(
+                query_text, model_name=qdrant_embed_model, expected_dim=qdrant_vector_size
+            )
+        except Exception as exc:  # pragma: no cover - runtime dependency
+            LOGGER.error("Embedding failed: %s", exc, exc_info=config.elastic.debug)
+            return jsonify({"error": "embedding_unavailable"}), 500
+
+        qdrant_vector_payload: Any
+        if qdrant_vector_name:
+            qdrant_vector_payload = {qdrant_vector_name: query_vector}
+        else:
+            qdrant_vector_payload = query_vector
+
+        qdrant_body: Dict[str, Any] = {
+            "vector": qdrant_vector_payload,
+            "limit": limit,
+            "offset": offset,
+            "with_payload": True,
+            "with_vectors": False,
+        }
+        if filters:
+            qdrant_body["filter"] = filters
+
+        try:
+            response = requests.post(
+                f"{qdrant_url}/collections/{qdrant_collection}/points/search",
+                json=qdrant_body,
+                timeout=20,
+            )
+            response.raise_for_status()
+            data = response.json()
+        except Exception as exc:
+            LOGGER.error("Vector search failed: %s", exc, exc_info=config.elastic.debug)
+            return jsonify({"error": "vector_search_unavailable"}), 502
+
+        points = data.get("result", []) if isinstance(data, dict) else []
+        items: List[Dict[str, Any]] = []
+        missing_channel_ids: Set[str] = set()
+        for point in points:
+            payload = point.get("payload", {}) or {}
+            raw_highlights = payload.get("highlights") or []
+            highlight_entries: List[Dict[str, str]] = []
+            for entry in raw_highlights:
+                if isinstance(entry, dict):
+                    html_value = entry.get("html") or entry.get("text")
+                else:
+                    html_value = str(entry)
+                if not html_value:
+                    continue
+                highlight_entries.append({"html": html_value, "source": "primary"})
+
+            channel_label = (
+                payload.get("channel_name")
+                or payload.get("channel_title")
+                or payload.get("channel_id")
+            )
+            items.append(
+                {
+                    "video_id": payload.get("video_id"),
+                    "channel_id": payload.get("channel_id"),
+                    "channel_name": channel_label,
+                    "title": payload.get("title"),
+                    "titleHtml": payload.get("title"),
+                    "description": payload.get("description"),
+                    "descriptionHtml": payload.get("description"),
+                    "date": payload.get("date"),
+                    "url": payload.get("url"),
+                    "chunkText": payload.get("text")
+                    or payload.get("chunk_text")
+                    or payload.get("chunk")
+                    or payload.get("content"),
+                    "chunkTimestamp": payload.get("timestamp")
+                    or payload.get("start_seconds")
+                    or payload.get("start"),
+                    "toHighlight": highlight_entries,
+                    "highlightSource": {
+                        "primary": bool(highlight_entries),
+                        "secondary": False,
+                    },
+                    "distance": point.get("score"),
+                    "internal_references_count": payload.get("internal_references_count", 0),
+                    "internal_references": payload.get("internal_references", []),
+                    "referenced_by_count": payload.get("referenced_by_count", 0),
+                    "referenced_by": payload.get("referenced_by", []),
+                    "video_status": payload.get("video_status"),
+                    "duration": payload.get("duration"),
+                }
+            )
+            if (not channel_label) and payload.get("channel_id"):
+                missing_channel_ids.add(str(payload.get("channel_id")))
+
+        if missing_channel_ids:
+            try:
+                es_lookup = client.search(
+                    index=index,
+                    body={
+                        "size": len(missing_channel_ids) * 2,
+                        "_source": ["channel_id", "channel_name"],
+                        "query": {"terms": {"channel_id.keyword": list(missing_channel_ids)}},
+                    },
+                )
+                hits = es_lookup.get("hits", {}).get("hits", [])
+                channel_lookup = {}
+                for hit in hits:
+                    src = hit.get("_source", {}) or {}
+                    cid = src.get("channel_id")
+                    cname = src.get("channel_name")
+                    if cid and cname and cid not in channel_lookup:
+                        channel_lookup[cid] = cname
+                for item in items:
+                    if not item.get("channel_name"):
+                        cid = item.get("channel_id")
+                        if cid and cid in channel_lookup:
+                            item["channel_name"] = channel_lookup[cid]
+            except Exception as exc:
+                LOGGER.debug("Vector channel lookup failed: %s", exc)
+
+        return jsonify(
+            {
+                "items": items,
+                "totalResults": len(items),
+                "offset": offset,
+            }
+        )
+
     @app.route("/api/transcript")
     def transcript():
         video_id = request.args.get("video_id", type=str)

static/frequency.html

@@ -4,6 +4,7 @@
     <meta charset="utf-8" />
     <meta name="viewport" content="width=device-width, initial-scale=1" />
     <title>Term Frequency Explorer</title>
+    <link rel="icon" href="/static/favicon.png" type="image/png" />
     <link rel="stylesheet" href="/static/style.css" />
     <style>
       #chart {
@@ -65,4 +66,3 @@
     <script src="/static/frequency.js"></script>
   </body>
 </html>
-

static/graph.html

@@ -0,0 +1,85 @@
+<!doctype html>
+<html>
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>TLC Reference Graph</title>
+    <link rel="icon" href="/static/favicon.png" type="image/png" />
+    <link rel="stylesheet" href="https://unpkg.com/xp.css" />
+    <link rel="stylesheet" href="/static/style.css" />
+    <script src="https://cdn.jsdelivr.net/npm/d3@7/dist/d3.min.js"></script>
+  </head>
+  <body>
+    <div class="window graph-window" style="max-width: 1100px; margin: 20px auto;">
+      <div class="title-bar">
+        <div class="title-bar-text">Reference Graph</div>
+        <div class="title-bar-controls">
+          <a class="title-bar-link" href="/">⬅ Search</a>
+        </div>
+      </div>
+      <div class="window-body">
+        <p>
+          Explore how videos reference each other. Enter a <code>video_id</code> to see its immediate
+          neighbors (referenced and referencing videos). Choose a larger depth to expand the graph.
+        </p>
+
+        <form id="graphForm" class="graph-controls">
+          <div class="field-group">
+            <label for="graphVideoId">Video ID</label>
+            <input
+              id="graphVideoId"
+              name="video_id"
+              type="text"
+              placeholder="e.g. dQw4w9WgXcQ"
+              required
+            />
+          </div>
+
+          <div class="field-group">
+            <label for="graphDepth">Depth</label>
+            <select id="graphDepth" name="depth">
+              <option value="1">1 hop</option>
+              <option value="2">2 hops</option>
+              <option value="3">3 hops</option>
+            </select>
+          </div>
+
+          <div class="field-group">
+            <label for="graphMaxNodes">Max nodes</label>
+            <select id="graphMaxNodes" name="max_nodes">
+              <option value="100">100</option>
+              <option value="150">150</option>
+              <option value="200" selected>200</option>
+              <option value="300">300</option>
+            </select>
+          </div>
+
+          <div class="field-group">
+            <label for="graphLabelSize">Labels</label>
+            <select id="graphLabelSize" name="label_size">
+              <option value="off">Off</option>
+              <option value="tiny" selected>Tiny</option>
+              <option value="small">Small</option>
+              <option value="normal">Normal</option>
+              <option value="medium">Medium</option>
+              <option value="large">Large</option>
+              <option value="xlarge">Extra large</option>
+            </select>
+          </div>
+
+          <button type="submit">Build graph</button>
+        </form>
+
+        <div id="graphStatus" class="graph-status">Enter a video ID to begin.</div>
+        <div id="graphContainer" class="graph-container"></div>
+      </div>
+
+      <div class="status-bar">
+        <p class="status-bar-field">Click nodes to open the video on YouTube</p>
+        <p class="status-bar-field">Colors represent channels</p>
+      </div>
+    </div>
+
+    <script src="/static/graph.js"></script>
+  </body>
+</html>

static/graph.js

@@ -0,0 +1,670 @@
+(() => {
+  const global = window;
+  const GraphUI = (global.GraphUI = global.GraphUI || {});
+  GraphUI.ready = false;
+  const form = document.getElementById("graphForm");
+  const videoInput = document.getElementById("graphVideoId");
+  const depthInput = document.getElementById("graphDepth");
+  const maxNodesInput = document.getElementById("graphMaxNodes");
+  const labelSizeInput = document.getElementById("graphLabelSize");
+  const statusEl = document.getElementById("graphStatus");
+  const container = document.getElementById("graphContainer");
+  const isEmbedded =
+    container && container.dataset && container.dataset.embedded === "true";
+
+  if (!form || !videoInput || !depthInput || !maxNodesInput || !labelSizeInput || !container) {
+    console.error("Graph: required DOM elements missing.");
+    return;
+  }
+
+  const color = d3.scaleOrdinal(d3.schemeTableau10);
+  const colorRange = typeof color.range === "function" ? color.range() : [];
+  const paletteSizeDefault = colorRange.length || 10;
+  const PATTERN_TYPES = [
+    { key: "none", legendClass: "none" },
+    { key: "diag-forward", legendClass: "diag-forward" },
+    { key: "diag-back", legendClass: "diag-back" },
+    { key: "cross", legendClass: "cross" },
+    { key: "dots", legendClass: "dots" },
+  ];
+  const ADDITIONAL_PATTERNS = PATTERN_TYPES.filter((pattern) => pattern.key !== "none");
+
+  const sanitizeDepth = (value) => {
+    const parsed = parseInt(value, 10);
+    if (Number.isNaN(parsed)) return 1;
+    return Math.max(0, Math.min(parsed, 3));
+  };
+
+  const sanitizeMaxNodes = (value) => {
+    const parsed = parseInt(value, 10);
+    if (Number.isNaN(parsed)) return 200;
+    return Math.max(10, Math.min(parsed, 400));
+  };
+
+  const LABEL_SIZE_VALUES = ["off", "tiny", "small", "normal", "medium", "large", "xlarge"];
+  const LABEL_FONT_SIZES = {
+    tiny: "7px",
+    small: "8px",
+    normal: "9px",
+    medium: "10px",
+    large: "11px",
+    xlarge: "13px",
+  };
+  const DEFAULT_LABEL_SIZE = "tiny";
+  const isValidLabelSize = (value) => LABEL_SIZE_VALUES.includes(value);
+
+  const getLabelSize = () => {
+    if (!labelSizeInput) return DEFAULT_LABEL_SIZE;
+    const value = labelSizeInput.value;
+    return isValidLabelSize(value) ? value : DEFAULT_LABEL_SIZE;
+  };
+
+  function setLabelSizeInput(value) {
+    if (!labelSizeInput) return;
+    labelSizeInput.value = isValidLabelSize(value) ? value : DEFAULT_LABEL_SIZE;
+  }
+
+  const getChannelLabel = (node) =>
+    (node && (node.channel_name || node.channel_id)) || "Unknown";
+
+  function appendPatternContent(pattern, baseColor, patternKey) {
+    pattern.append("rect").attr("width", 8).attr("height", 8).attr("fill", baseColor);
+
+    const strokeColor = "#1f1f1f";
+    const strokeOpacity = 0.35;
+
+    const addForward = () => {
+      pattern
+        .append("path")
+        .attr("d", "M-2,6 L2,2 M0,8 L8,0 M6,10 L10,4")
+        .attr("stroke", strokeColor)
+        .attr("stroke-width", 1)
+        .attr("stroke-opacity", strokeOpacity)
+        .attr("fill", "none");
+    };
+
+    const addBackward = () => {
+      pattern
+        .append("path")
+        .attr("d", "M-2,2 L2,6 M0,0 L8,8 M6,-2 L10,2")
+        .attr("stroke", strokeColor)
+        .attr("stroke-width", 1)
+        .attr("stroke-opacity", strokeOpacity)
+        .attr("fill", "none");
+    };
+
+    switch (patternKey) {
+      case "diag-forward":
+        addForward();
+        break;
+      case "diag-back":
+        addBackward();
+        break;
+      case "cross":
+        addForward();
+        addBackward();
+        break;
+      case "dots":
+        pattern
+          .append("circle")
+          .attr("cx", 4)
+          .attr("cy", 4)
+          .attr("r", 1.5)
+          .attr("fill", strokeColor)
+          .attr("fill-opacity", strokeOpacity);
+        break;
+      default:
+        break;
+    }
+  }
+
+  function createChannelStyle(label, baseColor, patternKey) {
+    const patternInfo =
+      PATTERN_TYPES.find((pattern) => pattern.key === patternKey) || PATTERN_TYPES[0];
+    return {
+      baseColor,
+      hatch: patternInfo ? patternInfo.key : "none",
+      legendClass: patternInfo ? patternInfo.legendClass : "none",
+    };
+  }
+
+  let currentGraphData = null;
+  let currentChannelStyles = new Map();
+  let currentDepth = sanitizeDepth(depthInput.value);
+  let currentMaxNodes = sanitizeMaxNodes(maxNodesInput.value);
+  let currentSimulation = null;
+
+  function setStatus(message, isError = false) {
+    if (!statusEl) return;
+    statusEl.textContent = message;
+    if (isError) {
+      statusEl.classList.add("error");
+    } else {
+      statusEl.classList.remove("error");
+    }
+  }
+
+  function sanitizeId(value) {
+    return (value || "").trim();
+  }
+
+  async function fetchGraph(videoId, depth, maxNodes) {
+    const params = new URLSearchParams();
+    params.set("video_id", videoId);
+    params.set("depth", String(depth));
+    params.set("max_nodes", String(maxNodes));
+    const response = await fetch(`/api/graph?${params.toString()}`);
+    if (!response.ok) {
+      const errorPayload = await response.json().catch(() => ({}));
+      const errorMessage =
+        errorPayload.error ||
+        `Graph request failed (${response.status} ${response.statusText})`;
+      throw new Error(errorMessage);
+    }
+    return response.json();
+  }
+
+  function resizeContainer() {
+    if (!container) return;
+    const minHeight = 520;
+    const viewportHeight = window.innerHeight;
+    container.style.height = `${Math.max(minHeight, Math.round(viewportHeight * 0.6))}px`;
+  }
+
+  function renderGraph(data, labelSize = "normal") {
+    if (!container) return;
+
+    if (currentSimulation) {
+      currentSimulation.stop();
+      currentSimulation = null;
+    }
+    container.innerHTML = "";
+
+    const width = container.clientWidth || 900;
+    const height = container.clientHeight || 600;
+
+    const svg = d3
+      .select(container)
+      .append("svg")
+      .attr("viewBox", [0, 0, width, height])
+      .attr("width", "100%")
+      .attr("height", height);
+
+    const defs = svg.append("defs");
+
+    defs
+      .append("marker")
+      .attr("id", "arrow-references")
+      .attr("viewBox", "0 -5 10 10")
+      .attr("refX", 18)
+      .attr("refY", 0)
+      .attr("markerWidth", 6)
+      .attr("markerHeight", 6)
+      .attr("orient", "auto")
+      .append("path")
+      .attr("d", "M0,-5L10,0L0,5")
+      .attr("fill", "#6c83c7");
+
+    defs
+      .append("marker")
+      .attr("id", "arrow-referenced-by")
+      .attr("viewBox", "0 -5 10 10")
+      .attr("refX", 18)
+      .attr("refY", 0)
+      .attr("markerWidth", 6)
+      .attr("markerHeight", 6)
+      .attr("orient", "auto")
+      .append("path")
+      .attr("d", "M0,-5L10,0L0,5")
+      .attr("fill", "#c76c6c");
+
+    const contentGroup = svg.append("g").attr("class", "graph-content");
+    const linkGroup = contentGroup.append("g").attr("class", "graph-links");
+    const nodeGroup = contentGroup.append("g").attr("class", "graph-nodes");
+    const labelGroup = contentGroup.append("g").attr("class", "graph-labels");
+
+    const links = data.links || [];
+    const nodes = data.nodes || [];
+
+    currentChannelStyles = new Map();
+    const uniqueChannels = [];
+    nodes.forEach((node) => {
+      const label = getChannelLabel(node);
+      if (!currentChannelStyles.has(label)) {
+        uniqueChannels.push(label);
+      }
+    });
+
+    const additionalPatternCount = ADDITIONAL_PATTERNS.length;
+    uniqueChannels.forEach((label, idx) => {
+      const baseColor = color(label);
+      let patternKey = "none";
+      if (idx >= paletteSizeDefault && additionalPatternCount > 0) {
+        const patternInfo =
+          ADDITIONAL_PATTERNS[(idx - paletteSizeDefault) % additionalPatternCount];
+        patternKey = patternInfo.key;
+      }
+      const style = createChannelStyle(label, baseColor, patternKey);
+      currentChannelStyles.set(label, style);
+    });
+
+    const linkSelection = linkGroup
+      .selectAll("line")
+      .data(links)
+      .enter()
+      .append("line")
+      .attr("stroke-width", 1.2)
+      .attr("stroke", (d) =>
+        d.relation === "references" ? "#6c83c7" : "#c76c6c"
+      )
+      .attr("stroke-opacity", 0.7)
+      .attr("marker-end", (d) =>
+        d.relation === "references" ? "url(#arrow-references)" : "url(#arrow-referenced-by)"
+      );
+
+    let nodePatternCounter = 0;
+    const nodePatternRefs = new Map();
+
+    const getNodeFill = (node) => {
+      const style = currentChannelStyles.get(getChannelLabel(node));
+      if (!style) {
+        return color(getChannelLabel(node));
+      }
+      if (!style.hatch || style.hatch === "none") {
+        return style.baseColor;
+      }
+      const patternId = `node-pattern-${nodePatternCounter++}`;
+      const pattern = defs
+        .append("pattern")
+        .attr("id", patternId)
+        .attr("patternUnits", "userSpaceOnUse")
+        .attr("width", 8)
+        .attr("height", 8);
+      appendPatternContent(pattern, style.baseColor, style.hatch);
+      pattern.attr("patternTransform", "translate(0,0)");
+      nodePatternRefs.set(node.id, pattern);
+      return `url(#${patternId})`;
+    };
+
+    const nodeSelection = nodeGroup
+      .selectAll("circle")
+      .data(nodes, (d) => d.id)
+      .enter()
+      .append("circle")
+      .attr("r", (d) => (d.is_root ? 10 : 7))
+      .attr("fill", (d) => getNodeFill(d))
+      .attr("stroke", "#1f1f1f")
+      .attr("stroke-width", (d) => (d.is_root ? 2 : 1))
+      .call(
+        d3
+          .drag()
+          .on("start", (event, d) => {
+            if (!event.active) simulation.alphaTarget(0.3).restart();
+            d.fx = d.x;
+            d.fy = d.y;
+          })
+          .on("drag", (event, d) => {
+            d.fx = event.x;
+            d.fy = event.y;
+          })
+          .on("end", (event, d) => {
+            if (!event.active) simulation.alphaTarget(0);
+            d.fx = null;
+            d.fy = null;
+          })
+      )
+      .on("click", (event, d) => {
+        if (d.url) {
+          window.open(d.url, "_blank", "noopener");
+        }
+      })
+      .on("contextmenu", (event, d) => {
+        event.preventDefault();
+        loadGraph(d.id, currentDepth, currentMaxNodes, { updateInputs: true });
+      });
+
+    nodeSelection
+      .append("title")
+      .text((d) => {
+        const parts = [];
+        parts.push(d.title || d.id);
+        if (d.channel_name) {
+          parts.push(`Channel: ${d.channel_name}`);
+        }
+        if (d.date) {
+          parts.push(`Date: ${d.date}`);
+        }
+        return parts.join("\n");
+      });
+
+    const labelSelection = labelGroup
+      .selectAll("text")
+      .data(nodes, (d) => d.id)
+      .enter()
+      .append("text")
+      .attr("class", "graph-node-label")
+      .attr("text-anchor", "middle")
+      .attr("fill", "#1f1f1f")
+      .attr("pointer-events", "none")
+      .text((d) => d.title || d.id);
+
+    applyLabelAppearance(labelSelection, labelSize);
+
+    const simulation = d3
+      .forceSimulation(nodes)
+      .force(
+        "link",
+        d3
+          .forceLink(links)
+          .id((d) => d.id)
+          .distance(120)
+          .strength(0.8)
+      )
+      .force("charge", d3.forceManyBody().strength(-320))
+      .force("center", d3.forceCenter(width / 2, height / 2))
+      .force(
+        "collide",
+        d3.forceCollide().radius((d) => (d.is_root ? 20 : 14)).iterations(2)
+      );
+
+    simulation.on("tick", () => {
+      linkSelection
+        .attr("x1", (d) => d.source.x)
+        .attr("y1", (d) => d.source.y)
+        .attr("x2", (d) => d.target.x)
+        .attr("y2", (d) => d.target.y);
+
+      nodeSelection.attr("cx", (d) => d.x).attr("cy", (d) => d.y);
+
+      labelSelection.attr("x", (d) => d.x).attr("y", (d) => d.y - (d.is_root ? 14 : 12));
+
+      nodeSelection.each(function (d) {
+        const pattern = nodePatternRefs.get(d.id);
+        if (pattern) {
+          const safeX = Number.isFinite(d.x) ? d.x : 0;
+          const safeY = Number.isFinite(d.y) ? d.y : 0;
+          pattern.attr("patternTransform", `translate(${safeX}, ${safeY})`);
+        }
+      });
+    });
+
+    const zoomBehavior = d3
+      .zoom()
+      .scaleExtent([0.3, 3])
+      .on("zoom", (event) => {
+        contentGroup.attr("transform", event.transform);
+      });
+
+    svg.call(zoomBehavior);
+    currentSimulation = simulation;
+  }
+
+  async function loadGraph(videoId, depth, maxNodes, { updateInputs = false } = {}) {
+    const sanitizedId = sanitizeId(videoId);
+    if (!sanitizedId) {
+      setStatus("Please enter a video ID.", true);
+      return;
+    }
+    const safeDepth = sanitizeDepth(depth);
+    const safeMaxNodes = sanitizeMaxNodes(maxNodes);
+
+    if (updateInputs) {
+      videoInput.value = sanitizedId;
+      depthInput.value = String(safeDepth);
+      maxNodesInput.value = String(safeMaxNodes);
+    }
+
+    setStatus("Loading graph…");
+    try {
+      const data = await fetchGraph(sanitizedId, safeDepth, safeMaxNodes);
+      if (!data.nodes || data.nodes.length === 0) {
+        setStatus("No nodes returned for this video.", true);
+        container.innerHTML = "";
+        currentGraphData = null;
+        currentChannelStyles = new Map();
+        renderLegend([]);
+        return;
+      }
+      currentGraphData = data;
+      currentDepth = safeDepth;
+      currentMaxNodes = safeMaxNodes;
+      renderGraph(data, getLabelSize());
+      renderLegend(data.nodes);
+      setStatus(
+        `Showing ${data.nodes.length} nodes and ${data.links.length} links (depth ${data.depth})`
+      );
+      updateUrlState(sanitizedId, safeDepth, safeMaxNodes, getLabelSize());
+    } catch (err) {
+      console.error(err);
+      setStatus(err.message || "Failed to build graph.", true);
+      container.innerHTML = "";
+      currentGraphData = null;
+      currentChannelStyles = new Map();
+      renderLegend([]);
+    }
+  }
+
+  async function handleSubmit(event) {
+    event.preventDefault();
+    await loadGraph(videoInput.value, depthInput.value, maxNodesInput.value, {
+      updateInputs: true,
+    });
+  }
+
+  function renderLegend(nodes) {
+    let legend = document.getElementById("graphLegend");
+    if (!legend) {
+      legend = document.createElement("div");
+      legend.id = "graphLegend";
+      legend.className = "graph-legend";
+      if (statusEl && statusEl.parentNode) {
+        statusEl.insertAdjacentElement("afterend", legend);
+      } else {
+        container.parentElement?.insertBefore(legend, container);
+      }
+    }
+
+    legend.innerHTML = "";
+
+    const edgesSection = document.createElement("div");
+    edgesSection.className = "graph-legend-section";
+
+    const edgesTitle = document.createElement("div");
+    edgesTitle.className = "graph-legend-title";
+    edgesTitle.textContent = "Edges";
+    edgesSection.appendChild(edgesTitle);
+
+    const createEdgeRow = (swatchClass, text) => {
+      const row = document.createElement("div");
+      row.className = "graph-legend-row";
+      const swatch = document.createElement("span");
+      swatch.className = `graph-legend-swatch ${swatchClass}`;
+      const label = document.createElement("span");
+      label.textContent = text;
+      row.appendChild(swatch);
+      row.appendChild(label);
+      return row;
+    };
+
+    edgesSection.appendChild(
+      createEdgeRow(
+        "graph-legend-swatch--references",
+        "Outgoing reference (video references other)"
+      )
+    );
+    edgesSection.appendChild(
+      createEdgeRow(
+        "graph-legend-swatch--referenced",
+        "Incoming reference (other video references this)"
+      )
+    );
+    legend.appendChild(edgesSection);
+
+    const channelSection = document.createElement("div");
+    channelSection.className = "graph-legend-section";
+    const channelTitle = document.createElement("div");
+    channelTitle.className = "graph-legend-title";
+    channelTitle.textContent = "Channels in view";
+    channelSection.appendChild(channelTitle);
+
+    const channelList = document.createElement("div");
+    channelList.className = "graph-legend-channel-list";
+
+    const channelEntries = Array.from(currentChannelStyles.entries()).sort((a, b) =>
+      a[0].localeCompare(b[0], undefined, { sensitivity: "base" })
+    );
+    const maxChannelItems = 20;
+
+    channelEntries.slice(0, maxChannelItems).forEach(([label, style]) => {
+      const item = document.createElement("div");
+      item.className = `graph-legend-channel graph-legend-channel--${
+        style.legendClass || "none"
+      }`;
+      const swatch = document.createElement("span");
+      swatch.className = "graph-legend-swatch graph-legend-channel-swatch";
+      swatch.style.backgroundColor = style.baseColor;
+      const text = document.createElement("span");
+      text.textContent = label;
+      item.appendChild(swatch);
+      item.appendChild(text);
+      channelList.appendChild(item);
+    });
+
+    const totalChannels = channelEntries.length;
+    if (channelList.childElementCount) {
+      channelSection.appendChild(channelList);
+      if (totalChannels > maxChannelItems) {
+        const note = document.createElement("div");
+        note.className = "graph-legend-note";
+        note.textContent = `+${totalChannels - maxChannelItems} more channels`;
+        channelSection.appendChild(note);
+      }
+    } else {
+      const empty = document.createElement("div");
+      empty.className = "graph-legend-note";
+      empty.textContent = "No channel data available.";
+      channelSection.appendChild(empty);
+    }
+
+    legend.appendChild(channelSection);
+  }
+
+  function applyLabelAppearance(selection, labelSize) {
+    if (labelSize === "off") {
+      selection.style("display", "none");
+    } else {
+      selection
+        .style("display", null)
+        .attr("font-size", LABEL_FONT_SIZES[labelSize] || LABEL_FONT_SIZES.normal);
+    }
+  }
+
+  function updateUrlState(videoId, depth, maxNodes, labelSize) {
+    if (isEmbedded) {
+      return;
+    }
+    const next = new URL(window.location.href);
+    next.searchParams.set("video_id", videoId);
+    next.searchParams.set("depth", String(depth));
+    next.searchParams.set("max_nodes", String(maxNodes));
+    if (labelSize && labelSize !== "normal") {
+      next.searchParams.set("label_size", labelSize);
+    } else {
+      next.searchParams.delete("label_size");
+    }
+    history.replaceState({}, "", next.toString());
+  }
+
+  function initFromQuery() {
+    const params = new URLSearchParams(window.location.search);
+    const videoId = sanitizeId(params.get("video_id"));
+    const depth = sanitizeDepth(params.get("depth") || "");
+    const maxNodes = sanitizeMaxNodes(params.get("max_nodes") || "");
+    const labelSizeParam = params.get("label_size");
+    if (videoId) {
+      videoInput.value = videoId;
+    }
+    depthInput.value = String(depth);
+    maxNodesInput.value = String(maxNodes);
+    if (labelSizeParam && isValidLabelSize(labelSizeParam)) {
+      setLabelSizeInput(labelSizeParam);
+    } else {
+      setLabelSizeInput(getLabelSize());
+    }
+    if (!videoId || isEmbedded) {
+      return;
+    }
+    loadGraph(videoId, depth, maxNodes, { updateInputs: false });
+  }
+
+  resizeContainer();
+  window.addEventListener("resize", resizeContainer);
+  form.addEventListener("submit", handleSubmit);
+  labelSizeInput.addEventListener("change", () => {
+    const size = getLabelSize();
+    if (currentGraphData) {
+      renderGraph(currentGraphData, size);
+      renderLegend(currentGraphData.nodes);
+    }
+    updateUrlState(
+      sanitizeId(videoInput.value),
+      currentDepth,
+      currentMaxNodes,
+      size
+    );
+  });
+  initFromQuery();
+
+  Object.assign(GraphUI, {
+    load(videoId, depth, maxNodes, options = {}) {
+      const targetDepth = depth != null ? depth : currentDepth;
+      const targetMax = maxNodes != null ? maxNodes : currentMaxNodes;
+      return loadGraph(videoId, targetDepth, targetMax, {
+        updateInputs: options.updateInputs !== false,
+      });
+    },
+    setLabelSize(size) {
+      if (!labelSizeInput || !size) return;
+      setLabelSizeInput(size);
+      labelSizeInput.dispatchEvent(new Event("change", { bubbles: true }));
+    },
+    setDepth(value) {
+      if (!depthInput) return;
+      const safe = sanitizeDepth(value);
+      depthInput.value = String(safe);
+      currentDepth = safe;
+    },
+    setMaxNodes(value) {
+      if (!maxNodesInput) return;
+      const safe = sanitizeMaxNodes(value);
+      maxNodesInput.value = String(safe);
+      currentMaxNodes = safe;
+    },
+    focusInput() {
+      if (videoInput) {
+        videoInput.focus();
+        videoInput.select();
+      }
+    },
+    stop() {
+      if (currentSimulation) {
+        currentSimulation.stop();
+        currentSimulation = null;
+      }
+    },
+    getState() {
+      return {
+        depth: currentDepth,
+        maxNodes: currentMaxNodes,
+        labelSize: getLabelSize(),
+        nodes: currentGraphData ? currentGraphData.nodes.slice() : [],
+        links: currentGraphData ? currentGraphData.links.slice() : [],
+      };
+    },
+    isEmbedded,
+  });
+  GraphUI.ready = true;
+  setTimeout(() => {
+    window.dispatchEvent(new CustomEvent("graph-ui-ready"));
+  }, 0);
+})();

static/index.html

@@ -3,7 +3,8 @@
   <head>
     <meta charset="utf-8" />
     <meta name="viewport" content="width=device-width, initial-scale=1" />
-    <title>This Little Corner (Python)</title>
+    <title>TLC Search</title>
+    <link rel="icon" href="/static/favicon.png" type="image/png" />
     <link rel="stylesheet" href="https://unpkg.com/xp.css" />
     <link rel="stylesheet" href="/static/style.css" />
     <script src="https://cdn.jsdelivr.net/npm/d3@7/dist/d3.min.js"></script>
@@ -11,8 +12,9 @@
   <body>
     <div class="window" style="max-width: 1200px; margin: 20px auto;">
       <div class="title-bar">
-        <div class="title-bar-text">This Little Corner — Elastic Search</div>
+        <div class="title-bar-text">This Little Corner</div>
         <div class="title-bar-controls">
+          <button id="aboutBtn" aria-label="About">?</button>
           <button id="minimizeBtn" aria-label="Minimize"></button>
           <button aria-label="Maximize"></button>
           <button aria-label="Close"></button>
@@ -20,6 +22,10 @@
       </div>
       <div class="window-body">
         <p>Enter a phrase to query title, description, and transcript text.</p>
+        <p style="font-size: 11px;">
+          Looking for semantic matches? Try the
+          <a href="/vector-search">vector search beta</a>.
+        </p>
 
         <fieldset>
           <legend>Search</legend>
@@ -30,19 +36,22 @@
           </div>
 
           <div class="field-row" style="margin-bottom: 8px; align-items: center;">
-            <label style="width: 60px;">Channel:</label>
+            <label for="channel" style="width: 60px;">Channel:</label>
-            <details id="channelDropdown" class="channel-dropdown" style="flex: 1;">
+            <select id="channel" style="flex: 1;">
-              <summary id="channelSummary">All Channels</summary>
+              <option value="">All Channels</option>
-              <div id="channelOptions" class="channel-options">
+            </select>
-                <div>Loading channels…</div>
+
-              </div>
+            <label for="year" style="margin-left: 8px;">Year:</label>
-            </details>
+            <select id="year">
+              <option value="">All Years</option>
+            </select>
 
             <label for="sort" style="margin-left: 8px;">Sort:</label>
             <select id="sort">
               <option value="relevant">Most relevant</option>
               <option value="newer">Newest first</option>
               <option value="older">Oldest first</option>
+              <option value="referenced">Most referenced</option>
             </select>
 
             <label for="size" style="margin-left: 8px;">Size:</label>
@@ -53,18 +62,30 @@
             </select>
           </div>
 
-          <div class="field-row">
+          <div class="field-row toggle-row">
-            <input type="checkbox" id="exactToggle" checked />
+            <div class="toggle-item toggle-item--first">
-            <label for="exactToggle">Exact</label>
+              <input type="checkbox" id="exactToggle" checked />
+              <label for="exactToggle">Exact</label>
+              <span class="toggle-help">Match all terms exactly.</span>
+            </div>
 
-            <input type="checkbox" id="fuzzyToggle" checked />
+            <div class="toggle-item">
-            <label for="fuzzyToggle">Fuzzy</label>
+              <input type="checkbox" id="fuzzyToggle" checked />
+              <label for="fuzzyToggle">Fuzzy</label>
+              <span class="toggle-help">Allow small typos and variations.</span>
+            </div>
 
-            <input type="checkbox" id="phraseToggle" checked />
+            <div class="toggle-item">
-            <label for="phraseToggle">Phrase</label>
+              <input type="checkbox" id="phraseToggle" checked />
+              <label for="phraseToggle">Phrase</label>
+              <span class="toggle-help">Boost exact phrases inside transcripts.</span>
+            </div>
 
-            <input type="checkbox" id="queryStringToggle" />
+            <div class="toggle-item">
-            <label for="queryStringToggle">Query string mode</label>
+              <input type="checkbox" id="queryStringToggle" />
+              <label for="queryStringToggle">Query string mode</label>
+              <span class="toggle-help">Use raw Lucene syntax (overrides other toggles).</span>
+            </div>
           </div>
         </fieldset>
 
@@ -78,7 +99,7 @@
             </fieldset>
           </div>
           <div class="summary-right">
-            <fieldset style="height: 100%;">
+            <fieldset>
               <legend>Timeline</legend>
               <div id="frequencySummary" style="font-size: 11px; margin-bottom: 8px;"></div>
               <div id="frequencyChart"></div>
@@ -92,11 +113,110 @@
         </fieldset>
       </div>
 
-      <div class="status-bar">
+    <div class="status-bar">
-        <p class="status-bar-field">Ready</p>
+      <p class="status-bar-field">Ready</p>
+    </div>
+    </div>
+
+        <div class="about-panel" id="aboutPanel" hidden>
+      <div class="about-panel__header">
+        <strong>About This App</strong>
+        <button id="aboutCloseBtn" aria-label="Close about panel">×</button>
+      </div>
+      <div class="about-panel__body">
+        <p>Use the toggles to choose exact, fuzzy, or phrase matching. Query string mode accepts raw Lucene syntax.</p>
+        <p>Results are ranked by your chosen sort order; the timeline summarizes the same query.</p>
+        <p>You can download transcripts, copy MLA citations, or explore references via the graph button.</p>
       </div>
     </div>
 
+<div
+      id="graphModalOverlay"
+      class="graph-modal-overlay"
+      aria-hidden="true"
+    >
+      <div
+        class="window graph-window graph-modal-window"
+        id="graphModalWindow"
+        role="dialog"
+        aria-modal="true"
+        aria-labelledby="graphModalTitle"
+      >
+        <div class="title-bar">
+          <div class="title-bar-text" id="graphModalTitle">Reference Graph</div>
+          <div class="title-bar-controls">
+            <button id="graphModalClose" aria-label="Close"></button>
+          </div>
+        </div>
+        <div class="window-body">
+          <p>
+            Explore how this video links with its neighbors. Adjust depth or node cap to expand the graph.
+          </p>
+
+          <form id="graphForm" class="graph-controls">
+            <div class="field-group">
+              <label for="graphVideoId">Video ID</label>
+              <input
+                id="graphVideoId"
+                name="video_id"
+                type="text"
+                placeholder="e.g. dQw4w9WgXcQ"
+                required
+              />
+            </div>
+
+            <div class="field-group">
+              <label for="graphDepth">Depth</label>
+              <select id="graphDepth" name="depth">
+                <option value="1" selected>1 hop</option>
+                <option value="2">2 hops</option>
+                <option value="3">3 hops</option>
+              </select>
+            </div>
+
+            <div class="field-group">
+              <label for="graphMaxNodes">Max nodes</label>
+              <select id="graphMaxNodes" name="max_nodes">
+                <option value="100">100</option>
+                <option value="150">150</option>
+                <option value="200" selected>200</option>
+                <option value="300">300</option>
+                <option value="400">400</option>
+              </select>
+            </div>
+
+            <div class="field-group">
+              <label for="graphLabelSize">Labels</label>
+              <select id="graphLabelSize" name="label_size">
+                <option value="off">Off</option>
+                <option value="tiny" selected>Tiny</option>
+                <option value="small">Small</option>
+                <option value="normal">Normal</option>
+                <option value="medium">Medium</option>
+                <option value="large">Large</option>
+                <option value="xlarge">Extra large</option>
+              </select>
+            </div>
+
+            <button type="submit">Build graph</button>
+          </form>
+
+          <div id="graphStatus" class="graph-status">Enter a video ID to begin.</div>
+          <div
+            id="graphContainer"
+            class="graph-container"
+            data-embedded="true"
+          ></div>
+        </div>
+
+        <div class="status-bar">
+          <p class="status-bar-field">Right-click a node to set a new root</p>
+          <p class="status-bar-field">Colors (and hatches) represent channels</p>
+        </div>
+      </div>
+    </div>
+
+    <script src="/static/graph.js"></script>
     <script src="/static/app.js"></script>
   </body>
 </html>

static/style.css

@@ -63,7 +63,7 @@ body.dimmed {
   }
 
   .field-row input[type="text"],
-  .field-row .channel-dropdown {
+  .field-row select#channel {
     flex: 1 1 100% !important;
     min-width: 0 !important;
     max-width: 100% !important;
@@ -86,63 +86,73 @@ body.dimmed {
     max-width: 100%;
     min-width: 100%;
   }
+
+  .graph-controls {
+    flex-direction: column;
+    align-items: stretch;
+  }
+
+  .graph-controls .field-group,
+  .graph-controls input,
+  .graph-controls select {
+    width: 100%;
+    min-width: 0;
+  }
 }
 
-/* Channel dropdown custom styling */
+.toggle-row {
-.channel-dropdown {
+  flex-direction: column;
-  position: relative;
+  align-items: flex-start;
-  display: inline-block;
+  gap: 4px;
+  margin-top: 8px;
 }
 
-.channel-dropdown summary {
+.toggle-row > * {
-  list-style: none;
+  margin-left: 0 !important;
-  cursor: pointer;
-  padding: 3px 4px;
-  background: ButtonFace;
-  border: 1px solid;
-  border-color: ButtonHighlight ButtonShadow ButtonShadow ButtonHighlight;
-  min-width: 180px;
-  text-align: left;
 }
 
-.channel-dropdown summary::-webkit-details-marker {
+.toggle-item {
-  display: none;
-}
-
-.channel-dropdown summary::after {
-  content: ' ▼';
-  font-size: 8px;
-  float: right;
-}
-
-.channel-dropdown[open] summary::after {
-  content: ' ▲';
-}
-
-.channel-options {
-  position: absolute;
-  margin-top: 2px;
-  padding: 4px;
-  background: ButtonFace;
-  border: 1px solid;
-  border-color: ButtonHighlight ButtonShadow ButtonShadow ButtonHighlight;
-  max-height: 300px;
-  overflow-y: auto;
-  box-shadow: 2px 2px 0 rgba(0, 0, 0, 0.2);
-  z-index: 100;
-  min-width: 220px;
-}
-
-.channel-option {
   display: flex;
   align-items: center;
   gap: 6px;
-  margin-bottom: 4px;
+  user-select: none;
-  font-size: 11px;
 }
 
-.channel-option:last-child {
+.toggle-item label {
-  margin-bottom: 0;
+  cursor: pointer;
+  width: auto !important;
+}
+
+.toggle-item--first {
+  margin-left: 0;
+}
+
+.toggle-item input[type="checkbox"] {
+  margin: 0;
+}
+
+.toggle-item input[type="checkbox"]:disabled + label {
+  color: GrayText;
+  opacity: 0.7;
+}
+
+.toggle-item input[type="checkbox"]:disabled {
+  cursor: not-allowed;
+}
+
+.toggle-item input[type="checkbox"]:disabled + label {
+  cursor: not-allowed;
+}
+
+.description-block {
+  background: Window;
+  border: 1px solid #919b9c;
+  padding: 6px 8px;
+  margin-top: 6px;
+  font-size: 11px;
+  white-space: pre-wrap;
+  max-height: 6em;
+  overflow-y: auto;
 }
 
 /* Layout helpers */
@@ -163,15 +173,373 @@ body.dimmed {
   min-width: 300px;
 }
 
+.graph-window {
+  width: 95%;
+}
+
+.graph-controls {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 12px;
+  align-items: flex-end;
+  margin-bottom: 12px;
+}
+
+.graph-controls .field-group {
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+}
+
+.graph-controls label {
+  font-size: 11px;
+  font-weight: bold;
+}
+
+.graph-controls input,
+.graph-controls select {
+  min-width: 160px;
+}
+
+.graph-status {
+  font-size: 11px;
+  margin-bottom: 8px;
+  color: #1f1f1f;
+}
+
+.graph-status.error {
+  color: #b00020;
+}
+
+.graph-container {
+  background: Window;
+  border: 1px solid #919b9c;
+  box-shadow: inset -1px -1px #0a0a0a, inset 1px 1px #fff;
+  position: relative;
+  width: 100%;
+  min-height: 520px;
+  height: auto;
+  overflow: visible;
+}
+
+.graph-modal-overlay {
+  position: fixed;
+  inset: 0;
+  display: none;
+  align-items: center;
+  justify-content: center;
+  padding: 24px;
+  background: rgba(0, 0, 0, 0.35);
+  z-index: 2000;
+}
+
+.graph-modal-overlay.active {
+  display: flex;
+}
+
+.graph-modal-window {
+  width: min(960px, 100%);
+  max-height: calc(100vh - 48px);
+}
+
+.graph-modal-window .window-body {
+  max-height: calc(100vh - 180px);
+  overflow-y: auto;
+}
+
+.graph-modal-window .graph-container {
+  height: 560px;
+}
+
+body.modal-open {
+  overflow: hidden;
+}
+
+.result-header {
+  display: flex;
+  justify-content: flex-start;
+  gap: 6px;
+  flex-wrap: wrap;
+  align-items: flex-start;
+}
+
+.result-header-main {
+  flex: 1 1 auto;
+  min-width: 220px;
+}
+
+.result-actions {
+  display: flex;
+  align-items: flex-start;
+  gap: 6px;
+  margin-left: auto;
+}
+
+.result-action-btn {
+  white-space: nowrap;
+  font-family: "Tahoma", "MS Sans Serif", sans-serif;
+  font-size: 11px;
+  padding: 4px 10px;
+}
+
+.result-meta {
+  display: flex;
+  align-items: center;
+  flex-wrap: wrap;
+  gap: 4px;
+}
+
+.result-status {
+  display: inline-flex;
+  align-items: center;
+  gap: 4px;
+  padding: 1px 6px;
+  border-radius: 3px;
+  font-size: 10px;
+  line-height: 1.3;
+  border: 1px solid #c4a3a3;
+  background: #fff6f6;
+  color: #6b1f1f;
+}
+
+.result-status::before {
+  content: "⚠";
+  font-size: 10px;
+  line-height: 1;
+}
+
+.result-status--deleted {
+  border-color: #d1a6a6;
+  background: #fff8f8;
+  color: #6b1f1f;
+}
+
+.graph-launch-btn {
+  white-space: nowrap;
+}
+
+.graph-node-label {
+  text-shadow: -1px -1px 0 #fff, 1px -1px 0 #fff, -1px 1px 0 #fff, 1px 1px 0 #fff;
+}
+
+.graph-nodes circle {
+  cursor: pointer;
+}
+
+.graph-legend {
+  margin: 12px 0;
+  font-size: 11px;
+  background: Window;
+  border: 1px solid #919b9c;
+  padding: 8px 10px;
+  display: inline-flex;
+  flex-direction: column;
+  gap: 4px;
+  box-shadow: inset -1px -1px #0a0a0a, inset 1px 1px #fff;
+}
+
+.graph-legend-section {
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+}
+
+.graph-legend-title {
+  font-weight: bold;
+  color: #1f1f1f;
+}
+
+.graph-legend-row {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.graph-legend-swatch {
+  display: inline-block;
+  width: 18px;
+  height: 12px;
+  border: 1px solid #1f1f1f;
+}
+
+.graph-legend-swatch--references {
+  background: #6c83c7;
+}
+
+.graph-legend-swatch--referenced {
+  background: #c76c6c;
+}
+
+.graph-legend-channel-list {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 8px;
+}
+
+.graph-legend-channel {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+}
+
+.graph-legend-channel-swatch {
+  width: 14px;
+  height: 14px;
+  background-repeat: repeat;
+  background-position: 0 0;
+  background-size: 6px 6px;
+}
+
+.graph-legend-channel--none .graph-legend-channel-swatch {
+  background-image: none;
+}
+
+.graph-legend-channel--diag-forward .graph-legend-channel-swatch {
+  background-image: repeating-linear-gradient(
+    45deg,
+    rgba(0, 0, 0, 0.35) 0,
+    rgba(0, 0, 0, 0.35) 2px,
+    transparent 2px,
+    transparent 4px
+  );
+  background-blend-mode: multiply;
+}
+
+.graph-legend-channel--diag-back .graph-legend-channel-swatch {
+  background-image: repeating-linear-gradient(
+    -45deg,
+    rgba(0, 0, 0, 0.35) 0,
+    rgba(0, 0, 0, 0.35) 2px,
+    transparent 2px,
+    transparent 4px
+  );
+  background-blend-mode: multiply;
+}
+
+.graph-legend-channel--cross .graph-legend-channel-swatch {
+  background-image:
+    repeating-linear-gradient(
+      45deg,
+      rgba(0, 0, 0, 0.25) 0,
+      rgba(0, 0, 0, 0.25) 2px,
+      transparent 2px,
+      transparent 4px
+    ),
+    repeating-linear-gradient(
+      -45deg,
+      rgba(0, 0, 0, 0.25) 0,
+      rgba(0, 0, 0, 0.25) 2px,
+      transparent 2px,
+      transparent 4px
+    );
+  background-blend-mode: multiply;
+}
+
+.graph-legend-channel--dots .graph-legend-channel-swatch {
+  background-image: radial-gradient(rgba(0, 0, 0, 0.35) 30%, transparent 31%);
+  background-size: 6px 6px;
+  background-blend-mode: multiply;
+}
+
+.graph-legend-note {
+  font-size: 10px;
+  color: #555;
+  font-style: italic;
+}
+
+.title-bar-link {
+  display: inline-block;
+  color: inherit;
+  text-decoration: none;
+  font-size: 11px;
+  padding: 2px 6px;
+  border: 1px solid;
+  border-color: ButtonHighlight ButtonShadow ButtonShadow ButtonHighlight;
+  background: ButtonFace;
+}
+
+.title-bar-controls #aboutBtn {
+  font-weight: bold;
+  font-size: 12px;
+  padding: 0 6px;
+  margin-right: 6px;
+}
+
+.toggle-item {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+}
+
+.toggle-help {
+  font-size: 10px;
+  color: #555;
+}
+
+.about-panel {
+  position: fixed;
+  top: 20px;
+  right: 20px;
+  width: 280px;
+  background: Window;
+  border: 2px solid #919b9c;
+  box-shadow: 0 4px 12px rgba(0, 0, 0, 0.25);
+  z-index: 2100;
+  font-size: 11px;
+}
+
+.about-panel__header {
+  display: flex;
+  justify-content: space-between;
+  align-items: center;
+  padding: 6px 8px;
+  background: #0055aa;
+  color: #fff;
+}
+
+.about-panel__body {
+  padding: 8px;
+  background: Window;
+  color: #000;
+}
+
+.about-panel__header button {
+  border: none;
+  background: transparent;
+  color: inherit;
+  font-weight: bold;
+  cursor: pointer;
+}
+
 /* Results styling */
 #results .item {
-  border-bottom: 1px solid ButtonShadow;
+  background: Window;
-  padding: 12px 0;
+  border: 2px solid #919b9c;
+  padding: 12px;
   margin-bottom: 8px;
+  max-width: 100%;
+  overflow: hidden;
+  word-wrap: break-word;
+  box-sizing: border-box;
+  box-shadow: 2px 2px 0 rgba(0, 0, 0, 0.15);
 }
 
 #results .item:last-child {
-  border-bottom: none;
+  margin-bottom: 0;
+}
+
+#results .item strong {
+  word-break: break-word;
+  max-width: 100%;
+  display: inline-block;
+}
+
+.window-body {
+  max-width: 100%;
+  overflow-x: hidden;
+  margin: 0;
+  padding: 1rem;
+  box-sizing: border-box;
 }
 
 /* Badges */
@@ -180,6 +548,8 @@ body.dimmed {
   display: flex;
   gap: 4px;
   flex-wrap: wrap;
+  max-width: 100%;
+  overflow: hidden;
 }
 
 .badge {
@@ -189,6 +559,25 @@ body.dimmed {
   padding: 2px 6px;
   font-size: 10px;
   font-weight: bold;
+  white-space: nowrap;
+  word-break: keep-all;
+}
+
+.badge--transcript-primary {
+  background: #0b6efd;
+}
+
+.badge--transcript-secondary {
+  background: #8f4bff;
+}
+
+.badge-clickable {
+  cursor: pointer;
+}
+
+.badge-clickable:focus {
+  outline: 2px solid rgba(11, 110, 253, 0.6);
+  outline-offset: 1px;
 }
 
 /* Transcript and highlights */
@@ -212,9 +601,14 @@ body.dimmed {
 }
 
 .highlight-row {
-  padding: 4px;
+  padding: 4px 6px;
   cursor: pointer;
   border: 1px solid transparent;
+  display: flex;
+  align-items: flex-start;
+  gap: 8px;
+  max-width: 100%;
+  box-sizing: border-box;
 }
 
 .highlight-row:hover {
@@ -223,6 +617,77 @@ body.dimmed {
   border: 1px dotted WindowText;
 }
 
+.highlight-text {
+  flex: 1 1 auto;
+  word-break: break-word;
+  overflow-wrap: anywhere;
+}
+
+.highlight-source-indicator {
+  width: 10px;
+  height: 10px;
+  border-radius: 2px;
+  border: 1px solid transparent;
+  margin-left: auto;
+  flex: 0 0 auto;
+}
+
+.highlight-source-indicator--primary {
+  background: #0b6efd;
+  border-color: #084bb5;
+}
+
+.highlight-source-indicator--secondary {
+  background: #8f4bff;
+  border-color: #5d2db3;
+}
+
+.vector-chunk {
+  margin-top: 8px;
+  padding: 8px;
+  background: #f3f7ff;
+  border: 1px solid #c7d0e2;
+  font-size: 11px;
+  line-height: 1.5;
+  word-break: break-word;
+}
+
+@media screen and (max-width: 640px) {
+  .result-header {
+    flex-direction: column;
+    gap: 6px;
+  }
+
+  .result-header-main {
+    flex: 1 1 auto;
+    min-width: 0;
+    width: 100%;
+  }
+
+  .result-actions {
+    width: auto;
+    align-self: flex-start;
+    justify-content: flex-start;
+    flex-wrap: wrap;
+    gap: 4px;
+    margin-left: 0;
+  }
+
+  .result-action-btn {
+    width: 100%;
+    text-align: left;
+  }
+
+  .highlight-row {
+    flex-direction: column;
+    gap: 4px;
+  }
+
+  .highlight-source-indicator {
+    align-self: flex-end;
+  }
+}
+
 mark {
   background: yellow;
   color: black;
@@ -237,8 +702,7 @@ mark {
   margin-top: 12px;
   padding: 8px;
   background: Window;
-  border: 2px solid;
+  border: 2px solid #919b9c;
-  border-color: ButtonShadow ButtonHighlight ButtonHighlight ButtonShadow;
   max-height: 400px;
   overflow-y: auto;
   font-size: 11px;
@@ -250,6 +714,10 @@ mark {
   border-bottom: 1px solid ButtonShadow;
 }
 
+.transcript-segment--matched {
+  background: #fff6cc;
+}
+
 .transcript-segment:last-child {
   border-bottom: none;
   margin-bottom: 0;
@@ -294,27 +762,9 @@ mark {
   line-height: 1.4;
 }
 
-.transcript-header {
+.transcript-header,
-  font-weight: bold;
-  margin-bottom: 8px;
-  display: flex;
-  align-items: center;
-  justify-content: space-between;
-  background: ActiveCaption;
-  color: CaptionText;
-  padding: 2px 4px;
-}
-
 .transcript-close {
-  cursor: pointer;
+  display: none;
-  font-size: 16px;
-  padding: 0 4px;
-  font-weight: bold;
-}
-
-.transcript-close:hover {
-  background: Highlight;
-  color: HighlightText;
 }
 
 /* Chart styling */

static/vector.html

@@ -0,0 +1,46 @@
+<!doctype html>
+<html>
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>TLC Vector Search</title>
+    <link rel="icon" href="/static/favicon.png" type="image/png" />
+    <link rel="stylesheet" href="https://unpkg.com/xp.css" />
+    <link rel="stylesheet" href="/static/style.css" />
+  </head>
+  <body>
+    <div class="window" style="max-width: 1200px; margin: 20px auto;">
+      <div class="title-bar">
+        <div class="title-bar-text">Vector Search (Experimental)</div>
+        <div class="title-bar-controls">
+          <a class="title-bar-link" href="/">⬅ Back to Search</a>
+        </div>
+      </div>
+      <div class="window-body">
+        <p>Enter a natural language prompt; results come from the Qdrant vector index.</p>
+
+        <fieldset>
+          <legend>Vector Query</legend>
+          <div class="field-row" style="margin-bottom: 8px;">
+            <label for="vectorQuery" style="width: 60px;">Query:</label>
+            <input id="vectorQuery" type="text" placeholder="Describe what you are looking for" style="flex: 1;" />
+            <button id="vectorSearchBtn">Search</button>
+          </div>
+        </fieldset>
+
+        <div id="vectorMeta" style="margin-top: 12px; font-size: 11px;"></div>
+
+        <fieldset style="margin-top: 16px;">
+          <legend>Results</legend>
+          <div id="vectorResults"></div>
+        </fieldset>
+      </div>
+
+      <div class="status-bar">
+        <p class="status-bar-field">Experimental mode • Qdrant</p>
+      </div>
+    </div>
+
+    <script src="/static/vector.js"></script>
+  </body>
+</html>

static/vector.js

@@ -0,0 +1,423 @@
+(() => {
+  const queryInput = document.getElementById("vectorQuery");
+  const searchBtn = document.getElementById("vectorSearchBtn");
+  const resultsDiv = document.getElementById("vectorResults");
+  const metaDiv = document.getElementById("vectorMeta");
+  const transcriptCache = new Map();
+
+  if (!queryInput || !searchBtn || !resultsDiv || !metaDiv) {
+    console.error("Vector search elements missing");
+    return;
+  }
+
+  /** Utility helpers **/
+  const escapeHtml = (str) =>
+    (str || "").replace(/[&<>"']/g, (ch) => {
+      switch (ch) {
+        case "&":
+          return "&amp;";
+        case "<":
+          return "&lt;";
+        case ">":
+          return "&gt;";
+        case '"':
+          return "&quot;";
+        case "'":
+          return "&#39;";
+        default:
+          return ch;
+      }
+    });
+
+  const fmtDate = (value) => {
+    try {
+      return (value || "").split("T")[0];
+    } catch {
+      return value;
+    }
+  };
+
+  const fmtSimilarity = (score) => {
+    if (typeof score !== "number" || Number.isNaN(score)) return "";
+    return score.toFixed(3);
+  };
+
+  const getVideoStatus = (item) =>
+    (item && item.video_status ? String(item.video_status).toLowerCase() : "");
+  const isLikelyDeleted = (item) => getVideoStatus(item) === "deleted";
+
+  const formatTimestamp = (seconds) => {
+    if (!seconds && seconds !== 0) return "00:00";
+    const hours = Math.floor(seconds / 3600);
+    const mins = Math.floor((seconds % 3600) / 60);
+    const secs = Math.floor(seconds % 60);
+    if (hours > 0) {
+      return `${hours}:${mins.toString().padStart(2, "0")}:${secs
+        .toString()
+        .padStart(2, "0")}`;
+    }
+    return `${mins}:${secs.toString().padStart(2, "0")}`;
+  };
+
+  const formatSegmentTimestamp = (segment) => {
+    if (!segment) return "";
+    if (segment.timestamp) return segment.timestamp;
+    const fields = [
+      segment.start_seconds,
+      segment.start,
+      segment.offset,
+      segment.time,
+    ];
+    for (const value of fields) {
+      if (value == null) continue;
+      const num = parseFloat(value);
+      if (!Number.isNaN(num)) {
+        return formatTimestamp(num);
+      }
+    }
+    return "";
+  };
+
+  const serializeTranscriptSection = (label, parts, fullText) => {
+    let content = "";
+    if (typeof fullText === "string" && fullText.trim()) {
+      content = fullText.trim();
+    } else if (Array.isArray(parts) && parts.length) {
+      content = parts
+        .map((segment) => {
+          const ts = formatSegmentTimestamp(segment);
+          const text = segment && segment.text ? segment.text : "";
+          return ts ? `[${ts}] ${text}` : text;
+        })
+        .join("\n")
+        .trim();
+    }
+    if (!content) return "";
+    return `${label}\n${content}\n`;
+  };
+
+  const fetchTranscriptData = async (videoId) => {
+    if (!videoId) return null;
+    if (transcriptCache.has(videoId)) {
+      return transcriptCache.get(videoId);
+    }
+    const res = await fetch(`/api/transcript?video_id=${encodeURIComponent(videoId)}`);
+    if (!res.ok) {
+      throw new Error(`Transcript fetch failed (${res.status})`);
+    }
+    const data = await res.json();
+    transcriptCache.set(videoId, data);
+    return data;
+  };
+
+  const buildTranscriptDownloadText = (item, transcriptData) => {
+    const lines = [];
+    lines.push(`Title: ${item.title || "Untitled"}`);
+    if (item.channel_name) lines.push(`Channel: ${item.channel_name}`);
+    if (item.date) lines.push(`Published: ${item.date}`);
+    if (item.url) lines.push(`URL: ${item.url}`);
+    lines.push("");
+
+    const primaryText = serializeTranscriptSection(
+      "Primary Transcript",
+      transcriptData.transcript_parts,
+      transcriptData.transcript_full
+    );
+    const secondaryText = serializeTranscriptSection(
+      "Secondary Transcript",
+      transcriptData.transcript_secondary_parts,
+      transcriptData.transcript_secondary_full
+    );
+
+    if (primaryText) lines.push(primaryText);
+    if (secondaryText) lines.push(secondaryText);
+    if (!primaryText && !secondaryText) {
+      lines.push("No transcript available.");
+    }
+    return lines.join("\n").trim() + "\n";
+  };
+
+  const flashButtonMessage = (button, message, duration = 1800) => {
+    if (!button) return;
+    const original = button.dataset.originalLabel || button.textContent;
+    button.dataset.originalLabel = original;
+    button.textContent = message;
+    setTimeout(() => {
+      button.textContent = button.dataset.originalLabel || original;
+    }, duration);
+  };
+
+  const handleTranscriptDownload = async (item, button) => {
+    if (!item.video_id) return;
+    button.disabled = true;
+    try {
+      const transcriptData = await fetchTranscriptData(item.video_id);
+      if (!transcriptData) throw new Error("Transcript unavailable");
+      const text = buildTranscriptDownloadText(item, transcriptData);
+      const blob = new Blob([text], { type: "text/plain" });
+      const url = URL.createObjectURL(blob);
+      const link = document.createElement("a");
+      link.href = url;
+      link.download = `${item.video_id}.txt`;
+      document.body.appendChild(link);
+      link.click();
+      document.body.removeChild(link);
+      URL.revokeObjectURL(url);
+      flashButtonMessage(button, "Downloaded");
+    } catch (err) {
+      console.error("Download failed", err);
+      alert("Unable to download transcript right now.");
+    } finally {
+      button.disabled = false;
+    }
+  };
+
+  const formatMlaDate = (value) => {
+    if (!value) return "n.d.";
+    const parsed = new Date(value);
+    if (Number.isNaN(parsed.valueOf())) return value;
+    const months = [
+      "Jan.", "Feb.", "Mar.", "Apr.", "May", "June",
+      "July", "Aug.", "Sept.", "Oct.", "Nov.", "Dec.",
+    ];
+    return `${parsed.getDate()} ${months[parsed.getMonth()]} ${parsed.getFullYear()}`;
+  };
+
+  const buildMlaCitation = (item) => {
+    const channel = (item.channel_name || item.channel_id || "Unknown").trim();
+    const title = (item.title || "Untitled").trim();
+    const url = item.url || "";
+    const publishDate = formatMlaDate(item.date);
+    const today = formatMlaDate(new Date().toISOString().split("T")[0]);
+    return `${channel}. "${title}." YouTube, uploaded by ${channel}, ${publishDate}, ${url}. Accessed ${today}.`;
+  };
+
+  const handleCopyCitation = async (item, button) => {
+    const citation = buildMlaCitation(item);
+    try {
+      if (navigator.clipboard && window.isSecureContext) {
+        await navigator.clipboard.writeText(citation);
+      } else {
+        const textarea = document.createElement("textarea");
+        textarea.value = citation;
+        textarea.style.position = "fixed";
+        textarea.style.opacity = "0";
+        document.body.appendChild(textarea);
+        textarea.select();
+        document.execCommand("copy");
+        document.body.removeChild(textarea);
+      }
+      flashButtonMessage(button, "Copied!");
+    } catch (err) {
+      console.error("Citation copy failed", err);
+      alert(citation);
+    }
+  };
+
+  /** Rendering helpers **/
+  const createHighlightRows = (entries) => {
+    if (!Array.isArray(entries) || !entries.length) return null;
+    const container = document.createElement("div");
+    container.className = "transcript highlight-list";
+    entries.forEach((entry) => {
+      if (!entry) return;
+      const row = document.createElement("div");
+      row.className = "highlight-row";
+      const textBlock = document.createElement("div");
+      textBlock.className = "highlight-text";
+      const html = entry.html || entry.text || entry;
+      textBlock.innerHTML = html || "";
+      row.appendChild(textBlock);
+      const indicator = document.createElement("span");
+      indicator.className = "highlight-source-indicator highlight-source-indicator--primary";
+      indicator.title = "Vector highlight";
+      row.appendChild(indicator);
+      container.appendChild(row);
+    });
+    return container;
+  };
+
+  const createActions = (item) => {
+    const actions = document.createElement("div");
+    actions.className = "result-actions";
+    const downloadBtn = document.createElement("button");
+    downloadBtn.type = "button";
+    downloadBtn.className = "result-action-btn";
+    downloadBtn.textContent = "Download transcript";
+    downloadBtn.addEventListener("click", () => handleTranscriptDownload(item, downloadBtn));
+    actions.appendChild(downloadBtn);
+
+    const citationBtn = document.createElement("button");
+    citationBtn.type = "button";
+    citationBtn.className = "result-action-btn";
+    citationBtn.textContent = "Copy citation";
+    citationBtn.addEventListener("click", () => handleCopyCitation(item, citationBtn));
+    actions.appendChild(citationBtn);
+
+    const graphBtn = document.createElement("button");
+    graphBtn.type = "button";
+    graphBtn.className = "result-action-btn graph-launch-btn";
+    graphBtn.textContent = "Graph";
+    graphBtn.disabled = !item.video_id;
+    graphBtn.addEventListener("click", () => {
+      if (!item.video_id) return;
+      const target = `/graph?video_id=${encodeURIComponent(item.video_id)}`;
+      window.open(target, "_blank", "noopener");
+    });
+    actions.appendChild(graphBtn);
+
+    return actions;
+  };
+
+  const renderVectorResults = (payload) => {
+    resultsDiv.innerHTML = "";
+    const items = payload.items || [];
+    if (!items.length) {
+      metaDiv.textContent = "No vector matches for this prompt.";
+      return;
+    }
+    metaDiv.textContent = `Matches: ${items.length} (vector mode)`;
+
+    items.forEach((item) => {
+      const el = document.createElement("div");
+      el.className = "item";
+      const header = document.createElement("div");
+      header.className = "result-header";
+      const headerMain = document.createElement("div");
+      headerMain.className = "result-header-main";
+      const titleEl = document.createElement("strong");
+      titleEl.innerHTML = item.titleHtml || escapeHtml(item.title || "Untitled");
+      headerMain.appendChild(titleEl);
+
+      const metaLine = document.createElement("div");
+      metaLine.className = "muted result-meta";
+      const channelLabel = item.channel_name || item.channel_id || "Unknown";
+      const dateLabel = fmtDate(item.date);
+      let durationSeconds = null;
+      if (typeof item.duration === "number") {
+        durationSeconds = item.duration;
+      } else if (typeof item.duration === "string" && item.duration.trim()) {
+        const parsed = parseFloat(item.duration);
+        if (!Number.isNaN(parsed)) {
+          durationSeconds = parsed;
+        }
+      }
+      const durationLabel = durationSeconds != null ? ` • ${formatTimestamp(durationSeconds)}` : "";
+      metaLine.textContent = channelLabel ? `${channelLabel} • ${dateLabel}${durationLabel}` : `${dateLabel}${durationLabel}`;
+      if (isLikelyDeleted(item)) {
+        metaLine.appendChild(document.createTextNode(" "));
+        const statusEl = document.createElement("span");
+        statusEl.className = "result-status result-status--deleted";
+        statusEl.textContent = "Likely deleted";
+        metaLine.appendChild(statusEl);
+      }
+      headerMain.appendChild(metaLine);
+
+      if (item.url) {
+        const linkLine = document.createElement("div");
+        linkLine.className = "muted";
+        const anchor = document.createElement("a");
+        anchor.href = item.url;
+        anchor.target = "_blank";
+        anchor.rel = "noopener";
+        anchor.textContent = "Open on YouTube";
+        linkLine.appendChild(anchor);
+        headerMain.appendChild(linkLine);
+      }
+
+      if (typeof item.distance === "number") {
+        const scoreLine = document.createElement("div");
+        scoreLine.className = "muted";
+        scoreLine.textContent = `Similarity score: ${fmtSimilarity(item.distance)}`;
+        headerMain.appendChild(scoreLine);
+      }
+
+      header.appendChild(headerMain);
+      header.appendChild(createActions(item));
+      el.appendChild(header);
+
+      if (item.descriptionHtml || item.description) {
+        const desc = document.createElement("div");
+        desc.className = "muted description-block";
+        desc.innerHTML = item.descriptionHtml || escapeHtml(item.description);
+        el.appendChild(desc);
+      }
+
+      if (item.chunkText) {
+        const chunkBlock = document.createElement("div");
+        chunkBlock.className = "vector-chunk";
+        if (item.chunkTimestamp && item.url) {
+          const tsObj =
+            typeof item.chunkTimestamp === "object"
+              ? item.chunkTimestamp
+              : { timestamp: item.chunkTimestamp };
+          const ts = formatSegmentTimestamp(tsObj);
+          const tsLink = document.createElement("a");
+          const paramValue =
+            typeof item.chunkTimestamp === "number"
+              ? Math.floor(item.chunkTimestamp)
+              : item.chunkTimestamp;
+          tsLink.href = `${item.url}${item.url.includes("?") ? "&" : "?"}t=${encodeURIComponent(
+            paramValue
+          )}`;
+          tsLink.target = "_blank";
+          tsLink.rel = "noopener";
+          tsLink.textContent = ts ? `[${ts}]` : "[timestamp]";
+          chunkBlock.appendChild(tsLink);
+          chunkBlock.appendChild(document.createTextNode(" "));
+        }
+        const chunkTextSpan = document.createElement("span");
+        chunkTextSpan.textContent = item.chunkText;
+        chunkBlock.appendChild(chunkTextSpan);
+        el.appendChild(chunkBlock);
+      }
+
+      const highlights = createHighlightRows(item.toHighlight);
+      if (highlights) {
+        el.appendChild(highlights);
+      }
+
+      resultsDiv.appendChild(el);
+    });
+  };
+
+  /** Search handler **/
+  const runVectorSearch = async () => {
+    const query = queryInput.value.trim();
+    if (!query) {
+      alert("Please enter a query.");
+      return;
+    }
+    metaDiv.textContent = "Searching vector index…";
+    resultsDiv.innerHTML = "";
+    searchBtn.disabled = true;
+    try {
+      const res = await fetch("/api/vector-search", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ query }),
+      });
+      if (!res.ok) {
+        throw new Error(`Vector search failed (${res.status})`);
+      }
+      const data = await res.json();
+      if (data.error) {
+        metaDiv.textContent = "Vector search unavailable.";
+        return;
+      }
+      renderVectorResults(data);
+    } catch (err) {
+      console.error(err);
+      metaDiv.textContent = "Vector search unavailable.";
+    } finally {
+      searchBtn.disabled = false;
+    }
+  };
+
+  searchBtn.addEventListener("click", runVectorSearch);
+  queryInput.addEventListener("keypress", (event) => {
+    if (event.key === "Enter") {
+      runVectorSearch();
+    }
+  });
+})();

sync_qdrant_channels.py

@@ -0,0 +1,188 @@
+"""
+Utility to backfill channel titles/names inside the Qdrant payloads.
+
+Usage:
+    python -m python_app.sync_qdrant_channels \
+        --batch-size 512 \
+        --max-batches 200 \
+        --dry-run
+"""
+
+from __future__ import annotations
+
+import argparse
+import logging
+from typing import Dict, Iterable, List, Optional, Set, Tuple
+import time
+
+import requests
+
+from .config import CONFIG
+from .search_app import _ensure_client
+
+LOGGER = logging.getLogger(__name__)
+
+
+def chunked(iterable: Iterable, size: int):
+    chunk: List = []
+    for item in iterable:
+        chunk.append(item)
+        if len(chunk) >= size:
+            yield chunk
+            chunk = []
+    if chunk:
+        yield chunk
+
+
+def resolve_channels(channel_ids: Iterable[str]) -> Dict[str, str]:
+    client = _ensure_client(CONFIG)
+    ids = list(set(channel_ids))
+    if not ids:
+        return {}
+    body = {
+        "size": len(ids) * 2,
+        "_source": ["channel_id", "channel_name"],
+        "query": {"terms": {"channel_id.keyword": ids}},
+    }
+    response = client.search(index=CONFIG.elastic.index, body=body)
+    resolved: Dict[str, str] = {}
+    for hit in response.get("hits", {}).get("hits", []):
+        source = hit.get("_source") or {}
+        cid = source.get("channel_id")
+        cname = source.get("channel_name")
+        if cid and cname and cid not in resolved:
+            resolved[cid] = cname
+    return resolved
+
+
+def upsert_channel_payload(
+    qdrant_url: str,
+    collection: str,
+    channel_id: str,
+    channel_name: str,
+    *,
+    dry_run: bool = False,
+) -> bool:
+    """Set channel_name/channel_title for all vectors with this channel_id."""
+    payload = {"channel_name": channel_name, "channel_title": channel_name}
+    body = {
+        "payload": payload,
+        "filter": {"must": [{"key": "channel_id", "match": {"value": channel_id}}]},
+    }
+    LOGGER.info("Updating channel_id=%s -> %s", channel_id, channel_name)
+    if dry_run:
+        return True
+    resp = requests.post(
+        f"{qdrant_url}/collections/{collection}/points/payload",
+        json=body,
+        timeout=120,
+    )
+    if resp.status_code >= 400:
+        LOGGER.error("Failed to update %s: %s", channel_id, resp.text)
+        return False
+    return True
+
+
+def scroll_missing_payloads(
+    qdrant_url: str,
+    collection: str,
+    batch_size: int,
+    *,
+    max_points: Optional[int] = None,
+) -> Iterable[List[Tuple[str, Dict[str, any]]]]:
+    """Yield batches of (point_id, payload) missing channel names."""
+    fetched = 0
+    next_page = None
+    while True:
+        current_limit = batch_size
+        while True:
+            body = {
+                "limit": current_limit,
+                "with_payload": True,
+                "filter": {"must": [{"is_empty": {"key": "channel_name"}}]},
+            }
+            if next_page:
+                body["offset"] = next_page
+            try:
+                resp = requests.post(
+                    f"{qdrant_url}/collections/{collection}/points/scroll",
+                    json=body,
+                    timeout=120,
+                )
+                resp.raise_for_status()
+                break
+            except requests.HTTPError as exc:
+                LOGGER.warning(
+                    "Scroll request failed at limit=%s: %s", current_limit, exc
+                )
+                if current_limit <= 5:
+                    raise
+                current_limit = max(5, current_limit // 2)
+                LOGGER.info("Reducing scroll batch size to %s", current_limit)
+                time.sleep(2)
+            except requests.RequestException as exc:  # type: ignore[attr-defined]
+                LOGGER.warning("Transient scroll error: %s", exc)
+                time.sleep(2)
+        payload = resp.json().get("result", {})
+        points = payload.get("points", [])
+        if not points:
+            break
+        batch: List[Tuple[str, Dict[str, any]]] = []
+        for point in points:
+            pid = point.get("id")
+            p_payload = point.get("payload") or {}
+            batch.append((pid, p_payload))
+        yield batch
+        fetched += len(points)
+        if max_points and fetched >= max_points:
+            break
+        next_page = payload.get("next_page_offset")
+        if not next_page:
+            break
+
+
+def main() -> None:
+    logging.basicConfig(level=logging.INFO, format="%(levelname)s %(message)s")
+    parser = argparse.ArgumentParser(
+        description="Backfill missing channel_name/channel_title in Qdrant payloads"
+    )
+    parser.add_argument("--batch-size", type=int, default=512)
+    parser.add_argument(
+        "--max-points",
+        type=int,
+        default=None,
+        help="Limit processing to the first N points for testing",
+    )
+    parser.add_argument("--dry-run", action="store_true")
+    args = parser.parse_args()
+
+    q_url = CONFIG.qdrant_url
+    collection = CONFIG.qdrant_collection
+    total_updates = 0
+
+    for batch in scroll_missing_payloads(
+        q_url, collection, args.batch_size, max_points=args.max_points
+    ):
+        channel_ids: Set[str] = set()
+        for _, payload in batch:
+            cid = payload.get("channel_id")
+            if cid:
+                channel_ids.add(str(cid))
+        if not channel_ids:
+            continue
+        resolved = resolve_channels(channel_ids)
+        if not resolved:
+            LOGGER.warning("No channel names resolved for ids: %s", channel_ids)
+            continue
+        for cid, name in resolved.items():
+            if upsert_channel_payload(
+                q_url, collection, cid, name, dry_run=args.dry_run
+            ):
+                total_updates += 1
+        LOGGER.info("Updated %s channel payloads so far", total_updates)
+
+    LOGGER.info("Finished. Total channel updates attempted: %s", total_updates)
+
+
+if __name__ == "__main__":
+    main()