README.md

metadata

title: Slack Bot ChromaDB
emoji: 🧠
colorFrom: blue
colorTo: green
sdk: docker
app_port: 8000
pinned: false
license: mit

ChromaDB for Slack Deno Bot

Hosts ChromaDB 1.5.9 for the Slack knowledge-base bot.
Public API: GET /api/v2/heartbeat

---

Hugging Face setup (exact steps)

1. Create the Space

1. Go to huggingface.co/new-space
2. Space name: e.g. slack-bot-chroma (becomes YOUR_USERNAME-slack-bot-chroma.hf.space)
3. License: MIT (or your choice)
4. SDK: Docker (not Gradio)
5. Hardware: CPU basic (free) is enough for demos; use persistent storage for production data
6. Click Create Space

2. Upload these two files to the Space repo

Copy from this folder in the bot repo:

File	Purpose
Dockerfile	Chroma 1.5.9 (listens on 8000 by default)
README.md	This file (app_port: 8000 tells HF which port to probe)

Option A — Git (recommended)

# Install HF CLI once: pip install huggingface_hub
huggingface-cli login

git clone https://huggingface.co/spaces/YOUR_USERNAME/slack-bot-chroma
cd slack-bot-chroma

# From your slack_deno_bot repo:
cp /path/to/slack_deno_bot/chromadb_server/huggingface/Dockerfile .
cp /path/to/slack_deno_bot/chromadb_server/huggingface/README.md .

git add Dockerfile README.md
git commit -m "Deploy ChromaDB 1.5.9 for Slack bot"
git push

Option B — Web UI

1. Open your Space → Files tab
2. Add file → upload Dockerfile and README.md (delete the auto-generated ones first if needed)
3. Space rebuilds automatically

3. Wait for the build

1. Open the Space → Logs tab
2. Wait until you see the container running (no build errors)
3. First boot can take 2–5 minutes on free tier

4. Enable persistent storage (recommended)

Without this, data is lost when the Space restarts or sleeps.

1. Space → Settings → Persistent storage
2. Turn on and pick a size (e.g. 5 GB for demos)
3. Data is stored at /data inside the container (Chroma default)

5. Verify the Space is live

Replace with your Space URL:

curl -sS https://YOUR_USERNAME-slack-bot-chroma.hf.space/api/v2/heartbeat

Expected: JSON response with "nanosecond heartbeat" (or similar), not an HTML error page.

Cold start: Free Spaces sleep after ~48h idle. The first request may take 30–60s while the Space wakes up.

6. Point the Slack bot project at Hugging Face

In your local slack_deno_bot/.env:

CHROMA_URL=https://YOUR_USERNAME-slack-bot-chroma.hf.space
CHROMA_COLLECTION=company_knowledge
GEMINI_API_KEY=your_key

Test from the bot repo:

deno task chroma:ping
deno task chroma:ingest -- knowledge_base_docs/example.md
deno task chroma:search -- "your test question"

7. Configure the deployed Slack app

Local .env is not used by slack deploy. Set prod env on the deployed app:

# Manifest outgoing domain (use your real hostname)
echo 'CHROMA_DEPLOY_URL=https://YOUR_USERNAME-slack-bot-chroma.hf.space' >> .env

slack env set CHROMA_URL "https://YOUR_USERNAME-slack-bot-chroma.hf.space" -a YOUR_DEPLOYED_APP_ID
slack deploy -a YOUR_DEPLOYED_APP_ID

8. Ingest documents into hosted Chroma

Run ingest from your laptop (embeddings use your GEMINI_API_KEY; vectors are sent to HF Chroma):

deno task chroma:ingest -- ./knowledge_base_docs -r

Re-run after adding or updating docs.

---

Space URL format

Space name	Public API base URL
slack-bot-chroma	https://YOUR_USERNAME-slack-bot-chroma.hf.space

Always use HTTPS, no port suffix (HF terminates TLS on 443).

---

Security notes

• Public Space: Anyone with the URL can call the Chroma API. Fine for internal demos; use a private HF Space or org access for production.
• Do not commit GEMINI_API_KEY or Slack secrets to the Space repo — only the Dockerfile and README belong there.
• Embeddings are computed locally/on Slack infra; HF only stores vectors and document text.

---

Troubleshooting

Problem	Fix
Space stuck on "Building" / slow start	Chroma runs on 8000 — README must have app_port: 8000, not 7860
Build fails	Check Logs; ensure sdk: docker and app_port: 8000 are in README YAML
502 / Space sleeping	Open the Space URL in a browser, wait, retry chroma:ping
Slack deploy domain_blocked	Set CHROMA_DEPLOY_URL in .env before slack deploy
Empty KB answers	Run chroma:ingest against the HF URL; confirm collection company_knowledge
Data gone after restart	Enable Persistent storage in Space settings