Skip to content

OpenJarvis User Manual

Intelligent AI Orchestrator with Multi-Model Routing — v0.1.0

OpenJarvis coordinates a team of specialized language models across any OpenAI-compatible API. Instead of relying on one model to handle everything, OpenJarvis routes each part of your conversation to dedicated specialists (such as code, mathematics, factual knowledge, and planning) and executes built-in tools when needed.

Why OpenJarvis?

Traditional AI assistants send all requests to a single model. OpenJarvis enables:

  • 🎯 Specialized Precision: Direct calculations to a math specialist, programming questions to a code specialist, and research to a knowledge specialist.
  • 🔌 Provider Freedom: Mix and match Ollama (free & private local models), OpenAI, Anthropic, Gemini, Groq, or custom OpenAI-compatible endpoints in one session.
  • 🛠️ 49 Production-Ready Built-In Tools: Automated tools for file inspection, Git/patch management, web requests, code execution, AST math solving, SQLite queries, datetime operations, and persistent memory.
  • 🪄 Zero-Fuss Onboarding: First-run interactive setup wizard creates your configuration automatically.
  • 🔍 Full Visibility: See every routing decision and tool invocation as your request is being processed.

Quick Example

$ openjarvis

oj> Calculate 15% of 80 and write a Python helper function for it

  ↳ routing: generalist → math
  ⚙ tool: evaluate_expression {"expression": "0.15 * 80"}
    → 12.0
  ↳ routing: math → code

15% of 80 is **12.0**.

Here is a Python function to calculate percentage values:

```python
def calculate_percentage(part_percent: float, total: float) -> float:
    """Calculate the given percentage of a total value."""
    return (part_percent / 100.0) * total
```

Key Capabilities

1. Multi-Model Orchestration

The generalist coordinates conversations, deciding whether to answer directly or route to a specialist. Specialists complete their domain task, call required tools, and return results to synthesize a final answer.

2. Built-In Tools Suite (49 Tools Across 9 Modules)

Category Count Highlight Capabilities
File I/O 9 Read/write files, recursive directory search, file regex match, glob locator, delete file
Git & Version Control 4 Unified git diff inspection, short status summary, commit log history, patch application
Web & API 5 Web search via DuckDuckGo, web page text extraction, Wikipedia lookups, REST HTTP, OpenAPI parsing
Code Execution 4 Sandboxed Python execution, shell commands, syntax linting, pytest test suite runner
Code Editor & Terminal 3 Atomic string replacement editor with undo history, bash command execution
Data Processing 6 JSON pretty-printing, jq queries, CSV table formatting, regex search/replace, SQLite querying
Math & Arithmetic 4 Safe expression evaluation, unit conversions, algebraic equation solving, prime factorization
Date & Time 4 ISO timestamps, timezone conversions, date arithmetic, days between dates
Session & Memory 10 Save/read/search persistent memories, session notes storage with backwards compatibility

3. Provider Agnostic

Configure each specialist to point to different endpoints, models, temperatures, and timeouts in your specialists.yaml file.

Next Steps

  1. Installation Guide → — Run via standalone pre-built binaries or from source
  2. Quick Start Guide → — Run the setup wizard and start chatting
  3. Configuration Overview → — Learn the specialists.yaml format
  4. Built-in Tools Manual → — Explore all 49 tools in detail
  5. Command-Line Usage → — Terminal shortcuts, CLI flags, and REPL commands