@amg555/frameforge-mcp 0.0.0-stage → 2.3.1

package/README.md

@@ -1,3 +1,378 @@
-# Temporary Holding Version
+<div align="center">
 
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.
+# ⚡ FrameForge
+
+### AI Video Production Engine for Any Agentic IDE
+
+**Any IDE. Any topic. Any style. 1080p videos from a single prompt.**
+
+[![npm](https://img.shields.io/npm/v/@amg555/frameforge-mcp.svg?label=npm)](https://www.npmjs.com/package/@amg555/frameforge-mcp)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+[![Node.js](https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen.svg)](https://nodejs.org/)
+[![Version](https://img.shields.io/badge/version-2.3.1-blue.svg)](CHANGELOG.md)
+[![MCP Spec](https://img.shields.io/badge/MCP-2024--11--05-blue.svg)](https://modelcontextprotocol.io/)
+[![Tests](https://img.shields.io/badge/tests-512%20passing-brightgreen.svg)](#testing)
+
+</div>
+
+---
+
+FrameForge is an MCP server your IDE agent calls to make videos. You (or your agent) write the scenes — FrameForge handles voice, images, FFmpeg compositing, and quality checking. Works in Bob, Cursor, Windsurf, Codex, opencode, Claude Code, and any MCP-capable IDE.
+
+---
+
+## Zero-install quickstart (npx)
+
+No cloning. No `npm install`. FFmpeg and edge-tts run on your own machine.
+
+```bash
+# 1. Install prerequisites (one-time, system-level)
+#    FFmpeg: https://ffmpeg.org/download.html  →  add bin/ to PATH
+#    edge-tts: pip install edge-tts
+
+# 2. Configure your API keys (one-time, saved to ~/.frameforge/config.json)
+npx --package=@amg555/frameforge-mcp frameforge-setup
+
+# 3. Check everything is ready
+npx --package=@amg555/frameforge-mcp frameforge-doctor
+
+# 4. Add to your IDE — then ask your agent to make a video
+```
+
+Add to your IDE's MCP config (works in Bob, Cursor, Windsurf, Codex, Claude Code, opencode):
+
+```json
+{
+  "mcpServers": {
+    "frameforge": {
+      "command": "npx",
+      "args": ["--package=@amg555/frameforge-mcp", "-y", "frameforge-mcp"]
+    }
+  }
+}
+```
+
+That's it. Your agent now has 17 video production tools. FFmpeg and TTS run on your machine — no cloud render costs.
+
+---
+
+## Other connection methods
+
+### Cloned repo (contributors / local dev)
+
+```bash
+git clone https://github.com/AMG555/FrameForge.git && cd FrameForge && npm run doctor && npm run setup
+```
+
+```json
+{
+  "mcpServers": {
+    "frameforge": {
+      "command": "node",
+      "args": ["/absolute/path/to/FrameForge/src/mcp-server/stdio.js"]
+    }
+  }
+}
+```
+
+### Remote URL (Cloudflare Workers — FFmpeg must run separately)
+
+Deploy once to Cloudflare (free):
+
+```bash
+cd cloudflare-mcp && npm install
+wrangler secret put WORKER_API_URL   # your FrameForge backend URL
+wrangler secret put MCP_AUTH_TOKEN   # any random 32-char string
+wrangler deploy
+```
+
+```json
+{
+  "mcpServers": {
+    "frameforge": {
+      "url": "https://video-mcp.YOUR_SUBDOMAIN.workers.dev/mcp",
+      "headers": { "Authorization": "Bearer YOUR_MCP_AUTH_TOKEN" }
+    }
+  }
+}
+```
+
+---
+
+## Make your first video
+
+**New to FrameForge?** Read [`START_HERE.md`](START_HERE.md) — it's a single file you paste into any IDE chat that starts a guided conversation: the agent asks what you want to make, produces a beat sheet for your approval, then drives the full pipeline.
+
+Once connected, tell your IDE:
+
+> *"Make a 60-second explainer video about how GPS works"*
+
+Or call the tool directly:
+
+```
+create_video_from_script({
+  scenes: [
+    {
+      narration: "GPS works by triangulating signals from at least four satellites.",
+      caption: "GPS TRIANGULATION",
+      visual_description: "satellite orbiting earth, flat illustration, bold colors"
+    },
+    {
+      narration: "Your phone measures tiny timing differences to pinpoint your location.",
+      caption: "YOUR LOCATION",
+      visual_description: "phone with pin on map, white background, minimal"
+    }
+  ],
+  title: "How GPS Works"
+})
+```
+
+Output lands in `.frameforge-files/`:
+
+```
+final_video_<id>.mp4   →  1080p landscape master
+short_1_<id>.mp4       →  auto-extracted 9:16 vertical Short
+video_meta.json        →  title, description, tags, chapters
+```
+
+**No API keys required for this path.** `edge-tts` (voice) and `pollinations` (images) are both free with no account. If Pollinations is slow, add `FALAI_API_KEY` and set `PROVIDER_IMAGE=falai`.
+
+---
+
+## Two ways to use it
+
+### Your IDE writes the script — no LLM key needed
+
+Your IDE's model writes the scenes. FrameForge only runs FFmpeg, TTS, and image generation. Use `create_video_from_script`. This is the fastest path and what most people want.
+
+### FrameForge researches and scripts itself
+
+For batch runs or when running without an IDE. Requires a free LLM key (Groq or Gemini both have free tiers).
+
+```bash
+npm run make -- --topic "How does GPS actually work?"
+```
+
+---
+
+## Bring your own media
+
+Skip AI generation for any scene by supplying your own files:
+
+```json
+{
+  "narration": "Here is the sorting algorithm running step by step.",
+  "caption": "ALGORITHM",
+  "video_path": "/recordings/screen_capture.mp4",
+  "audio_path": "/recordings/my_voice.mp3"
+}
+```
+
+Any combination of `image_path`, `video_path`, `audio_path` works — FrameForge only generates what you don't supply.
+
+---
+
+## Tools
+
+17 tools, grouped by what you're doing:
+
+**Make a video from your own script (no LLM key)**
+
+| Tool | What it does |
+|------|-------------|
+| `create_video_from_script` | Full render from scenes you provide — voice, images, FFmpeg, quality gate |
+| `ingest_media` | Import your own clips, images, and audio into a scenes manifest |
+| `edit_scenes` | Update captions, swap media files, or append scenes to a manifest |
+| `render_from_scenes` | Re-render from an existing manifest, skipping research and scripting |
+| `assemble_video` | FFmpeg render + quality gate — produces the master MP4 |
+| `render_tts` | Generate voice audio for one scene |
+| `multi_voice_tts` | Override TTS provider per call without touching `.env` |
+| `generate_scene_image` | Generate one scene image via the configured image provider |
+| `publish_video` | Deliver the finished video via file copy or HTTP webhook |
+| `external_platform_call` | HTTP bridge to ElevenLabs, Runway, Replicate, or any custom platform |
+
+**Auto-research pipeline (needs a free LLM key)**
+
+| Tool | What it does |
+|------|-------------|
+| `research_topic` | Research any topic and extract fact-checked claims |
+| `write_script` | Generate a structured scene script from research claims |
+| `queue_full_pipeline` | Queue a full research → render → publish run; returns `{ jobId }` immediately |
+| `get_job_status` | Poll a queued job for `{ status, stage, result }` |
+
+**Code-driven animated video (Playwright + FFmpeg)**
+
+| Tool | What it does |
+|------|-------------|
+| `scaffold_animation_project` | Create a project where every frame is a pure function of time `t` |
+| `validate_storyboard` | Check `STORYBOARD.md` against the 4-gate beat-sheet contract |
+| `render_animation` | Run the Playwright renderer: sheet / strip / storyboard / crop / full |
+
+---
+
+## Providers
+
+Switch any provider with one `.env` change — no code changes.
+
+**Voice (TTS)**
+
+| Provider | Free? | Key needed |
+|----------|-------|------------|
+| `edge-tts` | ✅ free, offline | none — **default** |
+| `elevenlabs` | paid | `ELEVENLABS_API_KEY` |
+| `azure` | paid | `AZURE_SPEECH_KEY` + `AZURE_SPEECH_REGION` |
+| `openai` | paid | `OPENAI_API_KEY` |
+| `webhook` | — | `WEBHOOK_TTS_URL` |
+
+Set `PROVIDER_TTS=<name>`.
+
+**Images**
+
+| Provider | Free? | Key needed |
+|----------|-------|------------|
+| `pollinations` | ✅ free, no account | none — **default** |
+| `falai` | ✅ free tier | `FALAI_API_KEY` at fal.ai |
+| `replicate` | paid | `REPLICATE_API_KEY` |
+| `dalle` | paid | `OPENAI_API_KEY` |
+| `webhook` | — | `WEBHOOK_IMAGE_URL` |
+
+Set `PROVIDER_IMAGE=<name>`.
+
+**LLM — only for the auto-research path**
+
+> Skip entirely when using `create_video_from_script`.
+
+| Provider | Free? | Key needed |
+|----------|-------|------------|
+| `groq` | ✅ free tier | `GROQ_API_KEY` at console.groq.com |
+| `gemini` | ✅ free tier | `GEMINI_API_KEY` at aistudio.google.com |
+| `ollama` | ✅ fully local | none — set `OLLAMA_MODEL` |
+| `openrouter` | pay per use | `OPENROUTER_API_KEY` |
+| `anthropic` | paid | `ANTHROPIC_API_KEY` |
+| `openai` | paid | `OPENAI_API_KEY` |
+
+Set `PROVIDER_LLM=<name>`.
+
+---
+
+## Quality gates
+
+Every video passes 7 checks before delivery. Any failure blocks the output.
+
+| Check | Requirement |
+|-------|------------|
+| Resolution | 1920×1080 landscape or 1080×1920 portrait |
+| Video codec | H.264, high profile, yuv420p |
+| Loudness | −14 LUFS ± 1.5 LU (broadcast standard) |
+| True peak | ≤ −0.5 dBTP |
+| Black frames | None longer than 0.5 s |
+| Frozen frames | None longer than 0.5 s |
+| File integrity | > 1 MB, valid video + audio streams |
+
+---
+
+## CLI commands
+
+**Via npx (no clone needed):**
+
+| Command | What it does |
+|---------|-------------|
+| `npx --package=@amg555/frameforge-mcp frameforge-setup` | Interactive key wizard — saves to `~/.frameforge/config.json` |
+| `npx --package=@amg555/frameforge-mcp frameforge-doctor` | Check Node, FFmpeg, edge-tts, and env vars |
+| `npx --package=@amg555/frameforge-mcp frameforge --topic "..."` | Full pipeline: research → render → deliver |
+
+**Via cloned repo:**
+
+| Command | What it does |
+|---------|-------------|
+| `npm run doctor` | Check Node, FFmpeg, edge-tts, and env vars |
+| `npm run setup` | Interactive key wizard — writes `.env` |
+| `npm run mcp` | Start MCP server on port 3210 |
+| `npm run mcp:stdio` | stdio MCP server for direct IDE connection |
+| `npm run make -- --topic "..."` | Full pipeline: research → render → deliver |
+| `npm run validate` | Deep structural validation (113 checks) |
+| `npm test` | Test suite (512 tests) |
+
+---
+
+## Architecture
+
+```
+YOUR IDE (Bob / Codex / opencode / Cursor / Windsurf / Claude Code)
+         │
+         │  HTTPS + SSE  (Cloudflare Remote MCP — optional)
+         ▼
+[Cloudflare Worker]     proxies tool calls, keeps IDE context lean
+         │
+         │  HTTP + INTERNAL_API_TOKEN
+         ▼
+[FrameForge Node]       runs on VPS / Pi / your machine / Render.com
+         │
+         ├─ src/mcp-server/     17 MCP tools
+         ├─ src/workflows/      research → script → assets → render → publish
+         ├─ src/pipeline/       state machine, retry, telemetry, manifest
+         ├─ src/video-renderer/ FFmpeg scene render + concat
+         ├─ src/audio-mixer/    LUFS normalise, sidechain ducking
+         └─ src/quality-gate/   7-point output validation
+```
+
+---
+
+## Security
+
+- **Timing-safe auth** — all token comparisons use `crypto.timingSafeEqual`
+- **SSRF protection** — webhook and pipeline URLs validated against RFC-1918 / loopback / link-local blocklist
+- **Path traversal guards** — `productionId`, `audioPath`, `localPath` validated before any file I/O
+- **CLI injection guards** — animation timestamps allowlisted before passing to `render.mjs`
+- **Worker API auth** — `INTERNAL_API_TOKEN` Bearer check on all internal routes
+- **Body size limits** — all HTTP servers reject requests > 4 MB with HTTP 413
+- **Credential redaction** — 14+ regex patterns strip API keys from all logs
+
+Report vulnerabilities: [SECURITY.md](SECURITY.md)
+
+---
+
+## Deployment
+
+**npx (no install — recommended for users):**
+
+```bash
+npx --package=@amg555/frameforge-mcp frameforge-setup
+# then add the IDE config from the quickstart above
+```
+
+**VPS / Pi / self-hosted (cloned repo):**
+
+```bash
+git clone https://github.com/AMG555/FrameForge.git && cd FrameForge
+cp docs/env.template .env   # edit with your keys
+npm run mcp
+```
+
+See [`docs/deployment.md`](docs/deployment.md) for Docker, systemd, and Render.com setup.
+
+---
+
+## Testing
+
+```bash
+npm test          # 512 tests, 0 failures
+npm run validate  # 113 structural checks
+```
+
+---
+
+## Contributing
+
+See [CONTRIBUTING.md](CONTRIBUTING.md).
+
+1. `npm run doctor` passes on your OS
+2. `npm test` passes (512/512)
+3. `npm run validate` passes (113/113)
+4. FFmpeg commands keep `-nostdin` and cross-platform font paths
+
+---
+
+## License
+
+MIT — see [LICENSE](LICENSE).

package/SECURITY.md

@@ -0,0 +1,71 @@
+# Security Policy
+
+## Supported Versions
+
+| Version | Supported |
+|:--- |:---: |
+| 2.x (current) | ✅ |
+| 1.x | ✅ security fixes only |
+| < 1.0 | ❌ |
+
+---
+
+## Reporting a Vulnerability
+
+**Please do not open a public GitHub Issue for security vulnerabilities.**
+
+Report security issues by emailing **amgt503@gmail.com** with:
+
+1. A description of the vulnerability and its potential impact.
+2. Steps to reproduce (minimal proof of concept if possible).
+3. Affected versions.
+4. Any suggested mitigation.
+
+You will receive an acknowledgement within **48 hours** and a resolution
+timeline within **7 days**. We will credit you in the release notes unless you
+prefer to remain anonymous.
+
+---
+
+## Security Model
+
+### Secrets handling
+- All credentials are loaded exclusively from environment variables (`.env`).
+- No secrets are committed to git — the `.gitignore` excludes `.env` and all
+  `*secrets*` files.
+- `src/utils/log-redaction.js` masks known credential patterns (API keys,
+  tokens, connection strings, Bearer tokens) from all log output.
+
+### Authentication
+- The Cloudflare Workers MCP endpoint requires a `Bearer` token
+  (`MCP_AUTH_TOKEN`) on every request.
+- The internal render API (`server-render-free.js`) requires
+  `INTERNAL_API_TOKEN` on all non-health routes.
+- Tokens are compared with `crypto.timingSafeEqual` to prevent timing attacks.
+
+### Shell injection prevention
+- All external process calls use `spawnSync(cmd, [args, as, array])` — never
+  template string interpolation into shell commands.
+
+### Network
+- FFmpeg calls use `-nostdin` to prevent interactive prompts from blocking.
+- All outbound HTTP calls (Pollinations, search APIs, webhook endpoints) use
+  Node.js `https` with explicit timeouts and no silent redirects beyond 5 hops.
+
+### Cloudflare Workers
+- The CF Workers MCP proxy does not store any credentials. `WORKER_API_URL`,
+  `MCP_AUTH_TOKEN`, and `INTERNAL_API_TOKEN` are set as Wrangler secrets
+  (encrypted at rest by Cloudflare).
+
+---
+
+## Known Limitations
+
+- The LLM fallback in `FactChecker._querySources()` produces synthesized
+  sources when no `SEARCH_API_KEY` is set. These are not real citations and
+  should not be presented as authoritative without a real search provider.
+- Pollinations.ai image requests are unauthenticated — do not include
+  private or sensitive content in image prompts.
+- `PUBLISH_WEBHOOK_URL` and `REVIEW_NOTIFY_URL` receivers are trusted
+  endpoints; ensure they are authenticated on your side before deploying
+  to production.

package/START_HERE.md

@@ -0,0 +1,200 @@
+# FrameForge — Start Here
+
+> **Paste this file's path into your agent IDE chat to start a guided video session.**
+> The agent will read this file, ask you what you want to make, and drive the full pipeline.
+
+---
+
+## For the Agent
+
+You are a video producer powered by FrameForge. Your job is to have a short conversation
+with the user to understand what video they want, then use the FrameForge MCP tools to
+produce it — from research through render to final MP4.
+
+**Do not assume anything. Ask first. Produce second.**
+
+---
+
+## Opening Message (say this verbatim to the user)
+
+> 👋 I'm your FrameForge video studio. I can make:
+>
+> - **Short-form reels** (Instagram, TikTok, YouTube Shorts — 15–60s, 9:16 portrait)
+> - **YouTube videos** (explainers, tutorials, long-form — 3–15 min, 16:9 landscape)
+> - **Code-driven animated explainers** (motion graphics, no stock images needed — any length)
+>
+> Let's start with a few quick questions:
+>
+> 1. **What's the topic or idea?** (e.g. "How black holes form", "5 tips for sourdough bread", "A travel reel about Tokyo")
+> 2. **Format?** `Reel / Short / TikTok (9:16)` · `YouTube video (16:9)` · `Animated explainer (any)`
+> 3. **Style?** `educational` · `comedic` · `documentary` · `cinematic` · `tutorial` · or describe your own
+> 4. **How long?** `Short (15–60s)` · `Medium (1–3 min)` · `Long (5–10 min)`
+> 5. **Voice?** `Auto-select` · or name a style: `professional narrator`, `friendly teacher`, `energetic host`
+> 6. **Any footage, images, or audio to include?** (paste file paths, or say "generate everything")
+>
+> Answer as many or as few as you like. I'll fill in the rest.
+
+---
+
+## After the User Answers — Agent Instructions
+
+Once you have the user's answers, choose the right render path and follow it precisely.
+
+---
+
+## Render Paths
+
+### PATH A — IDE-written scenes (fastest, most common)
+
+Use for: any video where you (the IDE agent) write the narration and visual descriptions.
+No LLM key needed — FrameForge generates TTS + AI images and renders with FFmpeg.
+
+**Step A1 — Beat Sheet Gate (MANDATORY)**
+
+Before calling any render tool, produce a beat sheet and **stop** — wait for approval:
+
+```
+BEAT SHEET — "<title>"
+Duration: Xs  |  Style: <style>  |  Format: <landscape/portrait>
+
+Scene 1  [0:00–0:08]  <what happens + what the viewer learns>  →  TRANSITION
+Scene 2  [0:08–0:18]  <what happens + what the viewer learns>  →  TRANSITION
+...
+Scene N  [X:XX–X:XX]  <what happens + what the viewer learns>  →  FADE OUT
+
+Voice: <voice description>
+Visual style: <1-sentence description>
+```
+
+Gate rules — revise if ANY are true:
+- A scene has no event (action that changes something)
+- Two scenes share the same timestamp range
+- Any seam is missing a transition label
+- Total duration doesn't match requested length ±10%
+
+**Step A2 — Render**
+
+```
+create_video_from_script({
+  scenes: [ { narration: "...", caption: "...", visual_description: "...", camera_hint: "zoom_in" } ],
+  title: "My Video",
+  orientation: "landscape",   // "portrait" for Reels/Shorts/TikTok
+  voice: "en-US-BrianNeural", // or use language: "en"
+  withMusic: true
+})
+```
+
+**Step A3 — Publish**
+
+Ask: "Save locally, or deliver somewhere?" Then call `publish_video`.
+
+---
+
+### PATH B — Code-driven animated video (motion graphics, explainers, reels)
+
+Use for: motion graphics, animated explainers, reels that must look high-quality without stock images.
+Requires one-time: `npm install playwright && npx playwright install chromium` (run in the project dir).
+
+**Step B1 — Beat sheet first (same gate as Path A — MANDATORY)**
+
+**Step B2 — TTS and transcription**
+
+```
+multi_voice_tts({ text: "Full narration.", outputPath: ".frameforge-files/voice_0.m4a" })
+// After scaffolding: python scripts/transcribe.py .frameforge-files/voice_0.m4a
+// → voice_0.captions.json — word timestamps for cuePhrase() sync
+```
+
+**Step B3 — Scaffold**
+
+```
+scaffold_animation_project({
+  outDir: ".frameforge-files/my-animation/",
+  title: "My Video",
+  width: 1920, height: 1080,   // 1080×1920 for portrait reels
+  scenes: [...],
+  audioPath: ".frameforge-files/voice_0.m4a"
+})
+```
+
+**Step B4 — Validate STORYBOARD.md**
+
+```
+validate_storyboard({ storyboardPath: ".frameforge-files/my-animation/STORYBOARD.md" })
+```
+
+**Step B5 — Write `edit.js`, then stills, then full render**
+
+```
+render_animation({ projectDir: "...", mode: "look" })           // pre-flight
+render_animation({ projectDir: "...", mode: "sheet" })          // one sheet per shot
+render_animation({ projectDir: "...", mode: "strip", timestamps: [2.1, 2.6] })  // motion check
+render_animation({ projectDir: "...", mode: "full", audioPath: ".frameforge-files/voice_0.m4a" })
+```
+
+**For reels: make 2–3 hook variants** (change only the first 3s of `edit.js`, full-render each):
+```
+render_animation({ projectDir: "...", mode: "stills", timestamps: [0.0, 1.5, 3.0] })
+// deliver hook_a.mp4, hook_b.mp4 so the user can A/B test
+```
+
+---
+
+### PATH C — Auto-research (needs `PROVIDER_LLM` + key in `.env`)
+
+Use for: overnight batch, topics where you want FrameForge to research and script everything.
+
+```
+queue_full_pipeline({ topic: "How GPS works", style: "educational" })
+→ { jobId }
+get_job_status({ jobId })   // poll every 30s
+```
+
+---
+
+## Quick Reference — MCP Tools
+
+| Tool | What it does |
+|------|-------------|
+| `research_topic` | Research any topic, extract verified claims |
+| `write_script` | Turn claims + beat sheet into a full narrated script |
+| `generate_scene_image` | AI-generated scene image from a prompt |
+| `render_tts` | Text-to-speech narration for a scene |
+| `create_video_from_script` | **IDE path** — you write the scenes, FrameForge renders |
+| `assemble_video` | Assemble rendered scenes into a master MP4 |
+| `render_animation` | Render HTML animation scenes (code-driven videos) |
+| `scaffold_animation_project` | Create a new code-driven animation project |
+| `ingest_media` | Import user-supplied images, clips, or audio |
+| `edit_scenes` | Re-render specific scenes with new prompts/captions |
+| `get_job_status` | Check async job progress |
+| `publish_video` | Upload to YouTube, push to webhook, or copy to folder |
+| `get_pipeline_status` | Health check and current job queue |
+| `external_platform_call` | Call any external API during production |
+
+Full tool docs: [`src/mcp-server/server.js`](src/mcp-server/server.js)
+
+---
+
+## Setup Check
+
+Before starting, confirm the server is running:
+
+```bash
+npm run doctor    # checks ffmpeg, fonts, TTS, LLM key
+npm run mcp       # starts the MCP server on port 3210
+```
+
+If `npm run doctor` reports missing dependencies, fix them before proceeding.
+See [`docs/env.template`](docs/env.template) for all environment variables.
+
+---
+
+## Determinism Note (for code-driven animations)
+
+If the user wants a **code-driven animation** (like the viral "Claude explains X" style),
+use `scaffold_animation_project` to create a project, then use `render_animation` to
+render it. Every frame is a pure function of time `t` — same input always produces the
+same pixels, enabling parallel render, resume, and diff-based review.
+
+See [`SKILL.md`](plugins/video-pipeline/skills/video-pipeline/SKILL.md) for the full
+code-driven animation workflow.

package/assets/metadata/assets.json

@@ -0,0 +1,5 @@
+{
+  "schema_version": "1.0.0",
+  "last_updated": "2026-09-09T19:45:00Z",
+  "assets": []
+}

package/assets/metadata/embeddings.json

@@ -0,0 +1,5 @@
+{
+  "schema_version": "1.0.0",
+  "last_updated": "2026-09-09T19:45:00Z",
+  "embeddings": {}
+}