vibe-coder

README

vibe-coder

vibe-coder is a local-first coding agent for Ollama, built in Go. It runs as a single static CLI binary and supports one-shot prompts, interactive REPL sessions, a rich tool system, session persistence with compaction, and optional RAG — all without leaving your machine.

Features

Core agent

• One-shot prompts (-p) and interactive REPL with streaming output.
• Multi-turn agent loop (up to 50 iterations, 2 retries) with tool observation feedback.
• Empty-response recovery — retries with escalating guidance when the model returns an empty reply.
• XML fallback parser — handles non-conformant LLMs that emit <invoke> blocks instead of native tool calls.
• Auto-parallel detection — recognizes 2-4 conjoined tasks in the user message and routes through ParallelAgents.

Tools (exposed to the model)

• File ops: Read, Write, Edit (with inline unified-diff preview in the TUI).
• Search: Glob, Grep (multiline, context lines, output modes).
• Shell: Bash and InteractiveBash with dangerous-command blocklists and protected-path guards.
• Web: WebFetch and WebSearch (DuckDuckGo scrape) with SSRF protection.
• Notebook: NotebookEdit for .ipynb JSON round-trip.
• Tasks: TodoWrite (live to-do panel), TaskStart, TaskList, TaskComplete, TaskCancel.
• Questions: AskUserQuestion for interactive multi-choice prompts.
• Orchestration: SubAgent (bounded fan-out) and ParallelAgents.
• All file-search tools skip heavy directories (.git, node_modules, vendor, etc.) and respect cancellation.

Session management

• Atomic persistence — append-only JSONL with temp-file + rename (0o600).
• Project-aware indexing — sessions are keyed by sha256(cwd)[:16] so --resume picks up where you left off per project.
• Compaction — triggered at 300 messages or 70 % of context window; a sidecar model summarizes the oldest messages while keeping the last 30 verbatim.
• Token estimate — maintained incrementally so compaction checks are O(1).

Safety and permissions

• Three permission tiers: Safe (always allow), Ask (prompt), Network (always prompt).
• -y mode flips global allow except for always-confirm Bash patterns.
• In-session memory — /yes and /no toggle at runtime; decisions are remembered for the current session.
• Persistent permissions — saved to <configDir>/permissions.json (skips Bash:allow).
• Dangerous-command blocklist (rm -rf /, > /dev/sda, etc.) and protected-path guard (/proc, /sys, ~/.ssh/id_*, ~/.aws/credentials, etc.).
• Environment scrubbing — Bash runs with safety.CleanEnv() to strip secrets.
• Plan mode — /plan restricts writes to <cwd>/.vibe-coder/plans/ until /approve returns to act mode.

TUI

• Two UI modes: plain (default, line-based) and rich (Bubble Tea + Lipgloss with pinned status bar, themed markdown, and syntax highlighting). Select with --ui rich.
• To-do panel — TodoWrite renders a live Cursor-style task list with status glyphs.
• Inline diff renderer — Edit tool shows colored unified-diff (red/green/cyan) with 50-line truncation.
• Type-ahead — keystrokes pressed mid-generation are captured and prefilled into the next prompt.
• Multiline input — dedicated keybinding starts multi-line mode; plain Enter submits.
• Terminal restoration — signal handler ensures raw mode is never left behind on Ctrl+C or panic.

Integrations

• RAG (optional, -tags rag) — SQLite-backed indexing and cosine-similarity retrieval. Build with --rag-index, query with --rag.
• MCP — stdio JSON-RPC client that discovers external tools and wraps them as mcp_<server>_<name>.
• Skills auto-load — searches three directories for skill markdown files (50 KiB cap, sanitized).
• Git checkpoint + auto-test — pre-Edit/Write stash; failing tests are re-injected as [AUTO-TEST] observations.
• File watcher — external file changes appear as [System Note] N file change(s) detected on the next iteration.

Requirements

• Go 1.25+
• A running Ollama instance for model-backed execution

Install

Quick install (one-liner)

Linux / macOS (requires curl):

curl -sL https://raw.githubusercontent.com/jonathanhecl/vibe-coder/main/install.sh | bash

Windows (requires PowerShell 5.1+):

irm https://raw.githubusercontent.com/jonathanhecl/vibe-coder/main/install.ps1 | iex

The scripts download the latest pre-built release for your OS/architecture from GitHub Releases, extract it, and place the binary in a standard location (~/.local/bin on Linux/macOS, %LOCALAPPDATA%\Programs\vibe-coder on Windows).

Install with Go

If you have Go installed:

go install github.com/jonathanhecl/vibe-coder/cmd/vibe-coder@latest

Make sure your GOBIN or GOPATH/bin is in PATH.

Manual download

Grab the archive for your platform from the GitHub Releases page, extract it, and move the binary to a directory in your PATH.

Build from source

go build -o vibe-coder ./cmd/vibe-coder

Windows:

go build -o vibe-coder.exe ./cmd/vibe-coder

Verify the install:

vibe-coder --version

Build

go build -o vibe-coder ./cmd/vibe-coder

Windows:

go build -o vibe-coder.exe ./cmd/vibe-coder

Helper scripts (PowerShell and Bash):

.\build.ps1      # dev build with timestamp + short hash
.\run.ps1        # build + run with forwarded flags
.\release.ps1    # cross-compile archives + checksums.txt

./build.sh
./run.sh
./release.sh

Quick Start

One-shot prompt:

./vibe-coder -p "Summarize this repository"

Interactive mode:

./vibe-coder

Use a specific model and host:

./vibe-coder --model llama3.1:8b --ollama-host http://127.0.0.1:11434

Send an initial prompt and keep chatting:

./vibe-coder -p "Refactor main.go" -i

Model Configuration

Model settings are loaded with this precedence:

1. defaults
2. config file
3. environment variables
4. CLI flags (highest priority)

Default config file path:

• Windows: %LOCALAPPDATA%\vibe-coder\vibe-coder.env
• Linux/macOS: ~/.config/vibe-coder/vibe-coder.env

You can override the config file path with:

• VIBE_CODER_CONFIG=<path>

Model keys and overrides:

• Config file key: MODEL=<model-name>
• Config file key: UI=plain|rich
• Config file key: SIDECAR_MODEL=<model-name>
• Environment: VIBE_CODER_MODEL=<model-name>
• Environment: VIBE_CODER_UI=plain|rich
• Environment: VIBE_CODER_SIDECAR_MODEL=<model-name>
• CLI: --ui plain|rich
• CLI: --model <model-name> (or -m <model-name>)

If no model is set, vibe-coder auto-selects one based on detected RAM tier.

Recommended model: ornith:9b works very well with Ollama for coding and multi-turn tool conversations in vibe-coder.

./vibe-coder --model ornith:9b --ollama-host http://127.0.0.1:11434

What is the sidecar model for?

MODEL is the conversational/coding model that answers every prompt. The sidecar is a smaller, faster model vibe-coder uses internally for short, high-leverage tasks the main model would either bloat the context with or answer too slowly. All sidecar calls are guarded by a worker semaphore, request deduplication (singleflight) and a small LRU cache, so even on a single local Ollama instance you never see N parallel requests piling up.

The sidecar is invoked in three places today:

1. Session compaction — when the session has more than 300 messages or the incremental token estimate exceeds 70% of ContextWindow, Session.Compact() sends the oldest messages to the sidecar with a "Summarize the conversation concisely" prompt and replaces them with the summary. The last 30 messages are kept verbatim.
2. Tool-output condensation — when a tool (typically Read, Bash, Grep) returns more than ~6 KB, the output is sent to the sidecar with a strict "produce 4-10 bullets, preserve paths/symbols/errors, no prose" system prompt. The condensed bullets replace the raw bytes in the model's context.
3. Path disambiguation — when the agent rescues a relative path (e.g. Read("config.go")) and finds multiple known absolute candidates, the sidecar picks one based on the user's current goal.

Pick a sidecar that is fast and cheap (e.g. llama3.2:3b, qwen3.5:4b, phi3:mini). Leave it empty to disable all three behaviours: compaction will truncate to a static "Earlier conversation truncated…" note, large tool outputs will be inserted verbatim into the context, and ambiguous paths will not be rescued.

Remote Ollama for vibe-coder only

If Ollama runs on another machine in your network, you can configure vibe-coder and persist those settings in one command, without changing global environment variables:

.\vibe-coder.exe -model "qwen3.5:9b" -sidecar "qwen3.5:4b" -ollama-host "http://192.168.1.50:11434" -save

What this does:

• Applies model, sidecar model, and host for the current run.
• Writes MODEL, SIDECAR_MODEL, and OLLAMA_HOST to %LOCALAPPDATA%\vibe-coder\vibe-coder.env.
• Keeps the change scoped to vibe-coder only (no setx needed).

Next runs can simply use:

.\vibe-coder.exe

If you use PowerShell and want to run from source with the same flags:

.\run.ps1 -model "qwen3.5:9b" -sidecar "qwen3.5:4b" -ollama-host "http://192.168.1.50:11434" -save

CLI Flags

• --version — print version and exit
• --help — show usage and exit
• --ui <mode> — UI mode (plain or rich)
• -p, --prompt <text> — one-shot prompt
• -i, --interactive — interactive mode (combine with -p to send an initial prompt and keep chatting)
• -m, --model <name> — model name
• --sidecar <name> — sidecar model name
• --no-sidecar — disable sidecar for this session only; with --save, persists SIDECAR_DISABLED=true
• -y, --yes — enable yes mode (auto-approve non-dangerous tools)
• --debug — enable debug logs
• --resume — resume the last session for this project
• --session-id <id> — resume a specific session
• --list-sessions — list known sessions
• --ollama-host <url> — Ollama base URL
• --max-tokens <n> — max generated tokens
• --temperature <f> — sampling temperature
• --context-window <n> — model context window
• --no-think — disable Ollama native thinking (faster replies)
• --hide-think — hide Ollama thinking blocks in CLI output
• --rag — enable RAG mode
• --rag-mode <type> — RAG mode type
• --rag-path <path> — RAG database path
• --rag-topk <n> — RAG top-k chunks
• --rag-model <name> — RAG embedding model
• --rag-index <path> — build/index RAG path and exit
• --save — persist MODEL, SIDECAR_MODEL, OLLAMA_HOST, and HIDE_THINK into vibe-coder.env

Slash Commands

Slash commands are entered at the > prompt during an interactive session.

Session

• /save — persist the current session to disk
• /clear — save and start a brand new session
• /sessions — list saved sessions (* = current project)
• /session <id> — resume a specific session quickly
• /sessions delete <id> — delete a specific session
• /sessions delete --all — delete every saved session
• /resume — resume the last session for this project path
• /resume <id> — resume a specific session by id (or unique prefix)
• /compact — force a sidecar-summarized compaction
• /tokens — show token usage vs the context window
• /status — show model, cwd, session and sidecar status

Model

• /model — show the active model
• /model <name> — switch the active model for this run
• /sidecar on|off — toggle the sidecar for this session
• /sidecar perm-on|perm-off — persist sidecar state to vibe-coder.env
• /sidecar status — show current sidecar state
• /hide-think — hide model thinking blocks in CLI output
• /show-think — show model thinking blocks in CLI output (default)

Mode

• /yes — auto-approve subsequent permission prompts
• /no — require manual approval (default)
• /plan — enter plan mode (writes restricted to .vibe-coder/plans/)
• /plan <goal> — enter plan mode and immediately start planning that goal
• /code — exit plan mode and return to coding mode
• /approve — exit plan mode and resume act mode in the same chat

Git

• /commit — stage + commit current changes (LLM-suggested message)

Misc

• /help — show the command reference
• /exit, /quit, /q — save and exit

Built-in Tool Notes

The agent exposes tools to the model through the system prompt. Users normally do not call these directly, but their behavior affects speed and context usage:

• Read accepts start_line, end_line, offset, limit, and max_bytes for partial file reads. Without those parameters it reads the full file with line numbers.
• Write creates or overwrites a file; dangerous paths and protected directories are blocked.
• Edit applies a replacement; the TUI renders a colored unified-diff preview.
• Glob accepts head_limit to bound large file listings.
• Grep accepts head_limit, offset, glob, output_mode, multiline, -i, -A, -B, and -C.
• Bash runs a shell command with cleaned environment and dangerous-pattern guards. InteractiveBash starts a persistent terminal session for multi-step CLI workflows.
• WebFetch and WebSearch fetch web content with SSRF protection.
• NotebookEdit edits .ipynb cells by JSON round-trip.
• TodoWrite maintains a live task list that the TUI renders as a panel.
• AskUserQuestion pauses the agent to ask the user a multi-choice question.
• SubAgent and ParallelAgents spawn child agents with bounded fan-out.
• Glob and Grep skip heavy directories such as .git, node_modules, vendor, dist, build, target, and .vibe-coder.
• Read, Glob, and Grep respect cancellation, so ESC/Ctrl-C can stop long file operations cleanly.

RAG Usage

Build an index:

./vibe-coder --rag-index ./somewhere

Run with RAG enabled:

./vibe-coder --rag -p "Find where permissions are enforced"

RAG is an optional build feature. To compile with RAG support:

go build -tags rag -o vibe-coder ./cmd/vibe-coder

Development

Running tests

# Default tests (no RAG)
go test ./...

# With RAG support
go test -tags rag ./...

Helper scripts

Script	Purpose
build.ps1 / build.sh	Dev build with timestamp + short Git hash
run.ps1 / run.sh	Build + run with forwarded CLI flags
release.ps1 / release.sh	Cross-compile for linux/amd64, linux/arm64, darwin/amd64, darwin/arm64, windows/amd64; produces archives + checksums.txt

Project layout

cmd/vibe-coder/          # Entry point and CLI wiring
internal/
  agent/                 # Agent loop, chat orchestration, tool execution
  config/                # Config loader (defaults < file < env < CLI)
  git/                   # Git checkpoint + auto-test
  mcp/                   # MCP stdio JSON-RPC client
  ollama/                # Ollama HTTP client (streaming NDJSON)
  onboarding/            # First-run interactive setup
  permissions/           # Permission tiers + in-session memory + persistence
  prompt/                # System prompt builder + instruction-file walk
  rag/                   # Optional SQLite RAG engine (build tag: rag)
  safety/                # Dangerous-command blocklist + env scrub
  session/               # Session persistence, compaction, project index
  sidecar/               # Sidecar model worker pool
  skills/                # Skill markdown auto-load
  slash/                 # Slash command dispatcher
  terminal/              # Interactive Bash session manager
  tools/                 # Built-in tool registry + implementations
  tui/                   # Plain and rich UI implementations
  version/               # Build-time version string
  watcher/               # File-change poller

Architecture Overview

vibe-coder is structured as a thin cmd/ layer over focused internal/ packages:

1. Config (internal/config) loads settings with the precedence defaults < vibe-coder.env < environment < CLI flags.
2. Agent (internal/agent) runs the multi-turn loop: chat once, parse tool calls, execute, feed results back, repeat. Capped at 50 iterations with 2 retries.
3. Session (internal/session) stores the transcript as append-only JSONL. It compacts automatically via the sidecar when the token estimate exceeds 70 % of the context window or message count exceeds 300.
4. Tools (internal/tools) register themselves into a central registry. The agent exposes their schemas to the model through the system prompt. MCP tools are discovered at startup and injected dynamically.
5. TUI (internal/tui) abstracts rendering behind a UI interface. PlainUI is the default; RichUI (Bubble Tea + Lipgloss) is selected with --ui rich.
6. Ollama client (internal/ollama) handles streaming NDJSON chat, model tags, version, and pull progress.

License

MIT