anveesa

A fast, multi-provider terminal AI coding assistant — interactive TUI, one-shot prompts, or web UI — backed by any OpenAI-compatible API.

What it is

Anveesa is a multi-provider terminal AI assistant that works with any OpenAI-compatible API — including OpenAI, Anthropic (via proxy), DeepSeek, Groq, Gemini, OpenRouter, local Ollama, and more. It ships a full TUI with streaming output, diff previews, cost tracking, and plan display, plus a lightweight web UI for browser-based access. You get 28 built-in tools covering file ops, git, web search, screenshot capture, and command execution — all with an approval flow that shows a full diff before any write.

Install

Cargo (recommended):

cargo install anveesa

Pre-built binaries:

Download from GitHub Releases for your platform (macOS, Linux x86_64/arm64).

npm wrapper:

npm install -g anveesa

Build from source:

git clone https://github.com/PandhuWibowo/anveesa-cli
cd anveesa-cli
cargo build --release
./target/release/anveesa

Requires Rust 1.85+ (2024 edition).

Quick start

# 1. Initialize config at ~/.config/anveesa/config.toml
anveesa config init

# 2. Edit config to set your provider and API key
#    (see sample below)

# 3. Set your API key in the environment
export OPENAI_API_KEY="sk-..."

# 4. Launch the TUI
anveesa

# 5. One-shot prompt
anveesa "explain the auth middleware in this repo"

# 6. Specific provider/model
anveesa --provider deepseek --model deepseek-chat "refactor src/auth.rs"

Sample config snippet (~/.config/anveesa/config.toml):

default_provider = "openai"

[providers.openai]
kind = "openai-compatible"
base_url = "https://api.openai.com/v1"
api_key_env = "OPENAI_API_KEY"
default_model = "gpt-4o"
fast_model = "gpt-4o-mini"

[providers.deepseek]
kind = "openai-compatible"
base_url = "https://api.deepseek.com"
api_key_env = "DEEPSEEK_API_KEY"
default_model = "deepseek-chat"

[providers.ollama]
kind = "openai-compatible"
base_url = "http://localhost:11434/v1"
default_model = "qwen2.5-coder:7b"

Providers

| Provider | kind | Notes | |---|---|---| | OpenAI | openai-compatible | GPT-4o, o1, o3 | | Anthropic (via proxy) | openai-compatible | Claude 3.5/3.7 via OpenRouter or direct | | DeepSeek | openai-compatible | deepseek-chat, deepseek-reasoner | | Groq | openai-compatible | Ultra-fast inference | | Google Gemini | openai-compatible | gemini-1.5-pro/flash | | OpenRouter | openai-compatible | 200+ models via one API | | Mistral | openai-compatible | mistral-large, codestral | | xAI (Grok) | openai-compatible | grok-2 | | Together AI | openai-compatible | Open-source models | | Fireworks AI | openai-compatible | Fast open-source | | Cerebras | openai-compatible | High-throughput | | SambaNova | openai-compatible | Enterprise inference | | NVIDIA NIM | openai-compatible | GPU-backed cloud | | GitHub Models | openai-compatible | Free tier models | | Perplexity | openai-compatible | Web-grounded answers | | Ollama | openai-compatible | Local models | | LM Studio | openai-compatible | Local GUI + API | | vLLM | openai-compatible | Self-hosted inference | | LiteLLM | openai-compatible | Universal proxy | | LocalAI | openai-compatible | Self-hosted | | Claude Code | command | Runs claude CLI | | Codex CLI | command | Runs codex CLI | | GitHub Copilot | command | Runs copilot CLI |

TUI commands

| Command | Description | |---|---| | /add <file> | Inject a file into context (Aider-style) | | /diff | Show git diff HEAD | | /commit [msg] | Stage all + commit; omit msg to have AI generate one | | /memory <note> | Append a note to .anveesa.md | | /clear | Reset conversation and session | | /undo | Restore last file changed by AI | | /compact | Drop old turns to free context window | | /copy | Copy last response to clipboard | | /export [path] | Save conversation as Markdown | | /model [name] | Show or switch model | | /provider [name] | Show or switch provider | | /status | Token usage and cost summary | | /search | Search conversation (or Ctrl+R) | | /help | Show all shortcuts | | /exit | Quit |

TUI keyboard shortcuts:

| Key | Action | |---|---| | Enter | Submit prompt | | Shift+Enter | Newline in input | | Tab | Complete /command, provider name, or file path | | Ctrl+R | Search conversation | | [ / ] | Navigate between file diffs and thinking blocks | | Enter (on diff) | Expand / collapse | | Ctrl+V / Cmd+V | Paste image or text | | j / k | Scroll (when input is empty) | | PageUp / PageDn | Scroll | | Ctrl+W | Delete word | | Ctrl+U | Clear line | | Ctrl+M | Toggle mouse capture (scroll vs. select mode) | | Ctrl+C | Cancel / quit |

Web UI

anveesa web           # starts on http://localhost:8374
anveesa web --port 8080

The web UI provides a browser-based chat interface with the same provider/model selection as the TUI. Features:

• Streaming responses with Markdown rendering
• Multi-provider switching
• Session history
• Cost and token tracking
• Mobile-friendly layout

Key features

• Multi-provider — 28 built-in providers; add any OpenAI-compatible endpoint in one config line
• Cost tracking — per-turn and session cost displayed in the header; custom pricing override per provider
• Session persistence — conversations auto-saved per working directory; resume where you left off
• Extended thinking — enable Anthropic extended thinking with extended_thinking = 10000 budget tokens
• Image attachments — paste screenshots with Cmd+V / Ctrl+V; queue multiple images per turn
• MCP support — connect Model Context Protocol servers for additional tools
• Aider-style /add — inject files into context without leaving the TUI
• Auto-compact — smart context compaction kicks in when the window fills up; AI summarises dropped turns
• Auto-commit helper — /commit stages all changes and pre-fills the input with an AI-generated commit message
• Model routing — fast_model for cheap read-only tool rounds; main model for final synthesis
• 28 built-in tools — file ops, git, web search, deep-fetch, screenshot, notes, run commands
• Approval flow — every write/run tool shows a diff preview before executing

Config reference

All fields of OpenAiCompatibleProviderConfig:

| Field | Type | Default | Description | |---|---|---|---| | base_url | string | required | API base URL (e.g. https://api.openai.com/v1) | | api_key | string | — | Inline API key (prefer api_key_env) | | api_key_env | string | — | Environment variable holding the API key | | default_model | string | — | Model to use when none is specified | | fast_model | string | — | Lightweight model for read-only tool rounds | | headers | table | {} | Extra HTTP headers (e.g. for GitHub Models) | | prompt_cache | bool | — | Enable prompt caching (Anthropic cache_control markers) | | max_tokens | int | — | Upper bound on tokens generated per response | | extended_thinking | int | — | Enable Anthropic extended thinking; value = budget_tokens | | pricing | [f64; 4] | — | Custom pricing [input, output, cache_read, cache_write] per million tokens |

Example with custom pricing:

[providers.my-provider]
kind = "openai-compatible"
base_url = "https://my-llm-api.com/v1"
api_key_env = "MY_API_KEY"
default_model = "my-model-v1"
pricing = [3.0, 15.0, 0.3, 3.75]

License

MIT — see LICENSE.