ResearchHelper CLI

ResearchHelper-CLI is an autonomous terminal-based research agent written in Go. It researches a thesis topic by iterating through a "Plan-Execute-Reflect" loop, utilizing internal tools (searching/scraping) and an integrated RAG (Retrieval-Augmented Generation) system for indexing knowledge.

Features

• Autonomous Research Loop:
  • Plan: Generates specific search queries using an LLM (Google Gemini).
  • Execute: Searches external sources (Arxiv, Brave, etc.) via internal Go tools.
  • Filter & Index: Filters results and indexes them into a vector database (PostgreSQL/pgvector).
  • Reflect: Decides whether to continue researching or finalize the report.
• Interactive & Non-Interactive Modes:
  • Run interactively with prompts.
  • Run via command-line flags for automation.
• Integrated RAG:
  • Directly indexes findings into the database for retrieval during the chat/report phase.
• Structured Outputs: Uses JSON Schema with the LLM to ensure reliable query generation.

Prerequisites

• Go 1.22+
• Make (optional, for build commands)
• PostgreSQL (with pgvector extension)

Configuration

The application is configured via environment variables. Create a .env file in the root directory:

# Required for the Main Agent Logic
GOOGLE_API_KEY=your_gemini_api_key

# Required for Research Tools
BRAVE_API_TOKEN=your_brave_search_token
MISTRAL_API_KEY=your_mistral_api_key
ANTHROPIC_API_KEY=your_anthropic_key # If used by other tools

# Database Configuration
DB_HOST=localhost
DB_PORT=5432
DB_USER=postgres
DB_PASSWORD=postgres
DB_NAME=research_agent

Installation & Build

1. Clone the repository:

   git clone https://github.com/mikeboe/research-helper.git
   cd research-helper

2. Build the CLI:

   make build

   This creates the binary at bin/research-helper.

3. Build the Server:

   make build-server

   This creates the binary at bin/research-server.

Usage

1. Interactive Mode

Simply run the binary without arguments. You will be prompted for the topic and the collection name.

./bin/research-helper

2. Non-Interactive Mode (Flags)

Use flags to automate the process.

./bin/research-helper --topic "Impact of LoRA on LLM Fine-tuning" --collection "thesis_v1"

• --topic, -t: The research topic (required in non-interactive mode).
• --collection, -c: The target RAG collection name (defaults to "thesis_db").

Development

• Run Tests: make test
• Clean: make clean
• Run Server: make server
• Run Web UI: make web