GoCP — Go Copilot with RAG & Tool Calling

GoCP is a Go-based AI copilot that works with Ollama LLMs, supporting automatic tool calling and RAG (Retrieval-Augmented Generation) for querying PDFs and live data — all from the terminal.

✨ Features

🤖 Terminal-based AI chat with Ollama

🔧 Automatic tool calling (Wikipedia, web search, crypto prices)

📄 RAG support for querying PDF documents

🔍 DuckDuckGo web search

💰 Real-time crypto prices via CoinGecko

🧠 Context pruning & conversation memory

🎯 Interactive model selection at startup

🏗️ Architecture

GoCP/
├── server/   # Tool execution & APIs (port 8080)
└── client/   # CLI chat client with RAG

Server: Tool registry, execution engine, APIs

Client: Chat UI, Ollama interaction, RAG pipeline

📋 Prerequisites

• Go 1.21+
• Ollama (running locally)

Recommended Models:

ollama pull llama3.1:8b
ollama pull qwen2.5:7b
ollama pull nomic-embed-text   # Required for RAG

🚀 Installation

git clone <repo-url>
cd GoCP

# Build Server
cd server
go mod tidy
go build -o gocp-server

# Build Client
cd ../client
go mod tidy
go build -o gocp-client

⚙️ Configuration

server/.env

PORT=8080

client/.env

OLLAMA_API_URL=http://localhost:11434
SERVER_URL=http://localhost:8080

▶️ Running GoCP

Terminal 1 — Server

cd server
./gocp-server

Terminal 2 — Client

cd client
./gocp-client

You'll be prompted to select a model and can start chatting immediately.

💡 Usage Examples

Ask Questions (Auto Tool Calling)

GoCP> What happened in tech today?
[🔧 Tool Called: web search]

Crypto Prices

GoCP> What's the Bitcoin price?
[🔧 Tool Called: Crypto]

Web Search

GoCP> Tell me about Alan Turing
[🔧 Tool Called: web search]

PDF RAG

GoCP> rag it
[📄 RAG Context Retrieved]

To Exit

GoCP> exit

🛠️ Available Tools

Tool	Purpose
fetching_wikipedia	Encyclopedic knowledge
fetching_crypto	Live crypto prices
fetching_online	Web search
RAG	Semantic search over PDFs

Tools are automatically selected by the model.

➕ Adding a New Tool (Quick Overview)

1. Define schema in server/schema/tools.json
2. Implement logic in server/tool_internals/
3. Add wrapper in server/tools/
4. Register in server/registery/
5. Restart server → tool becomes available

🐛 Troubleshooting

Ollama not responding

ollama serve
curl http://localhost:11434/api/tags

Server not reachable

curl http://localhost:8080/health

RAG not working

ollama pull nomic-embed-text

🧠 Performance Tips

• Use 8B models for best tool-calling
• Ensure at least 8–16GB RAM
• RAG context is auto-pruned for efficiency

🔮 Futures (Roadmap)

Planned and upcoming enhancements for GoCP:

📈 Stocks, Commodities & Metals

• Local symbol registry (stocks, ETFs, metals)
• Asset name → symbol resolution (no hallucination)
• Futures contracts support (e.g. GC=F, CL=F)
• Exchange-aware resolution (NASDAQ, NYSE, NSE)

🧠 Smarter RAG

• Multi-PDF indexing
• Chunk relevance scoring
• Persistent document memory per session
• Hybrid RAG (local + web)

🧰 Tooling Improvements

• Tool routing enforcement at code level
• Tool confidence scoring
• Pluggable tool backends
• Rate-limit aware fallbacks

⚡ Performance & UX

• Streaming responses
• Better context pruning
• Session save / restore
• Configurable memory depth

🛡️ Reliability & Safety

• Zero-hallucination symbol policy
• Explicit clarification on ambiguity
• Deterministic tool execution

🤝 Contributing

1. Fork the repo
2. Create a feature branch
3. Add code + tests
4. Open a PR 🚀

---

Built with ❤️ using Go & Ollama

Last updated: Jan 28, 2026