🐾 PAWSitiveOps – AnythingLLM

PAWSitiveOps-AnythingLLM is the AI backbone for the PAWSitiveOps platform.
It packages an AnythingLLM deployment and integration glue so we can run specialized compliance & operations chatbots for pet/vet centers, then expose them to React/FastAPI apps and n8n automations.

This repo focuses on infrastructure + integration (not the upstream AnythingLLM code).
Think: environment setup, deployment configs, routing, and recipes to wire agents into real workflows.

---

🎯 Purpose

• Provide a repeatable deployment of AnythingLLM (cloud or Docker)
• Standardize workspace/agent setup for PAWSitiveOps personas:
  • Compliance Hound
  • Sam the Safety Dog
  • Operations Bulldog
• Offer clean endpoints so frontends (React) and services (FastAPI, n8n) can talk to our agents
• Keep credentials and provider keys securely managed with environment variables

---

👨‍💻 My Role

• Designed the deployment approach (Render/Docker) and env layout
• Built integration patterns for webhooks, document ingestion, and chat completions
• Planned the multi-agent routing so different PAWSitiveOps bots can serve multiple apps
• Documented usage for teammates and future contributors

---

🛠️ Technologies Used

• AnythingLLM – LLM router, vector index, workspaces
• Docker / Docker Compose – Containerized deploys
• Render – (Planned/Active) one-click cloud hosting
• FastAPI – (Planned) thin API that fronts agent calls for apps
• React – (Planned) frontends consuming agent endpoints
• n8n – Workflow automation to orchestrate intake → embed → respond
• PostgreSQL / SQLite – Vector/storage options depending on target env
• Object Storage – (Optional) document archives (S3-compatible)

---

📦 Current & Planned Use Cases

✅ Current

• Stand up an AnythingLLM instance for PAWSitiveOps
• Define agent/workspace conventions (naming, roles, retrieval settings)
• Set environment scaffolding for LLM provider keys and embeddings

🚀 Planned

• Compliance Hound – Answers licensing/permit questions and links citations
• Sam the Safety Dog – Safety/SOP Q&A with document retrieval
• Operations Bulldog – Day-to-day ops, customer-service scripts, checklists
• Marketing Poodle (Optional) – Social copy & campaign ideas
• External Apps – VagalFit, DealAgent007, IndieSmithy UI consuming the same agent API
• n8n Pipelines – Auto-ingest PDFs/URLs → embed → notify → serve in chat

---

🔍 Workflow Example – Multi-App Agent Routing

[Frontend (React)] [n8n Webhook] [CLI/Script] \ | / \ | / \ | / +-----[ FastAPI Router ]-----+ | v [ AnythingLLM API ] | Workspaces | | Retrieval | | Embeddings | | [ Vector / Storage ] | v [ Response to Caller ]

---

⚙️ Quick Start

1) Environment Variables

Create .env with:

OPENAI_API_KEY=...
ANTHROPIC_API_KEY=...
GROQ_API_KEY=...
EMBEDDINGS_PROVIDER=openai
DATABASE_URL=sqlite:///./anythingllm.db
PORT=3001
CORS_ORIGIN=https://your-frontend.example


---

## ⚙️ Quick Start

### 1) Environment Variables
Create `.env` with:

```ini
OPENAI_API_KEY=...
ANTHROPIC_API_KEY=...
GROQ_API_KEY=...
EMBEDDINGS_PROVIDER=openai
DATABASE_URL=sqlite:///./anythingllm.db
PORT=3001
CORS_ORIGIN=https://your-frontend.example

version: "3.9"
services:
  anythingllm:
    image: mintplexlabs/anythingllm:latest
    ports:
      - "3001:3001"
    env_file:
      - .env
    volumes:
      - ./data:/app/server/storage
    restart: unless-stopped

Run:

docker compose up -d

3) Create Workspaces
pawsops-compliance-hound

pawsops-safety-dog

pawsops-operations-bulldog

Attach documents/links → set retrieval settings.

🐍 FastAPI Router Stub
Example main.py for routing chat calls to AnythingLLM:
from fastapi import FastAPI, HTTPException
import requests
import os

app = FastAPI()
ANYTHINGLLM_URL = os.getenv("ANYTHINGLLM_URL", "http://localhost:3001")
API_KEY = os.getenv("ANYTHINGLLM_API_KEY")

@app.post("/api/ask/{workspace}")
def ask_workspace(workspace: str, query: dict):
    headers = {"Authorization": f"Bearer {API_KEY}"}
    response = requests.post(
        f"{ANYTHINGLLM_URL}/api/v1/workspace/{workspace}/chat",
        headers=headers,
        json={"message": query.get("message", "")}
    )
    if response.status_code != 200:
        raise HTTPException(status_code=response.status_code, detail=response.text)
    return response.json()

📒 Workspace Playbook
A step-by-step guide for creating, maintaining, and updating workspaces is available in the Workspace Playbook.
It covers:

Naming conventions

Role descriptions

Retrieval settings

Document ingestion workflows (manual + n8n automated)

Best practices for prompt engineering in AnythingLLM

🔐 Security Notes
Keep provider keys in .env or Render Secrets

Restrict CORS to trusted domains

Use JWT or API key auth for public endpoints

Log requests for audit and debugging

📚 Lessons Learned
Thin router layer keeps apps clean

Workspace standardization avoids confusion

Automated ingestion is key for freshness

Prompts as config = easier iteration

🗺️ Future Plans
One-click Render deploy (AnythingLLM + FastAPI)

Workspace seeding script

Unified metrics/logging dashboard

React chatbot widget for embedding in sites

Enterprise connectors (SharePoint, S3, GDrive)