\n\n\n\n\nA lightweight, effective **(AST-based)** semantic code search tool for your codebase. Built on [CocoIndex](https://github.com/cocoindex-io/cocoindex) β a Rust-based ultra performant data transformation engine. Use it from the CLI, or integrate with Claude, Codex, Cursor β any coding agent β via [Skill](#skill-recommended) or [MCP](#mcp-server).\n\n- Instant token saving by 70%.\n- **1 min setup** β install and go, zero config needed!\n\n
\n\n[](https://discord.com/invite/zpA9S2DR7s)\n[](https://github.com/cocoindex-io/cocoindex)\n[](https://cocoindex.io/docs/getting_started/quickstart)\n[](https://opensource.org/licenses/Apache-2.0)\n\n[](https://pepy.tech/projects/cocoindex)\n[](https://github.com/cocoindex-io/cocoindex/actions/workflows/CI.yml)\n[](https://github.com/cocoindex-io/cocoindex/actions/workflows/release.yml)\n\n\nπ Please help star [CocoIndex](https://github.com/cocoindex-io/cocoindex) if you like this project!\n\n[Deutsch](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=de) |\n[English](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=en) |\n[EspaΓ±ol](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=es) |\n[franΓ§ais](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=fr) |\n[ζ₯ζ¬θͺ](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=ja) |\n[νκ΅μ΄](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=ko) |\n[PortuguΓͺs](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=pt) |\n[Π ΡΡΡΠΊΠΈΠΉ](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=ru) |\n[δΈζ](https://readme-i18n.com/cocoindex-io/cocoindex-code?lang=zh)\n\n
\n\n\n## Get Started β zero config, let's go!\n\n### Install\n\nUsing [pipx](https://pipx.pypa.io/stable/installation/):\n```bash\npipx install 'cocoindex-code[full]' # batteries included (local embeddings)\npipx upgrade cocoindex-code # upgrade\n```\n\nUsing [uv](https://docs.astral.sh/uv/getting-started/installation/):\n```bash\nuv tool install --upgrade 'cocoindex-code[full]'\n```\n\nTwo install styles β they mirror the Docker image variants of the same names:\n- `cocoindex-code[full]` β batteries-included. Pulls in `sentence-transformers` so local embeddings (no API key required) work out of the box. The `ccc init` interactive prompt defaults to [Snowflake/snowflake-arctic-embed-xs](https://huggingface.co/Snowflake/snowflake-arctic-embed-xs).\n- `cocoindex-code` (slim) β LiteLLM-only; requires a cloud embedding provider and API key. Use when you don't want the local-embedding deps (~1 GB of torch + transformers).\n\nNext, set up your [coding agent integration](#coding-agent-integration) β or jump to [Manual CLI Usage](#manual-cli-usage) if you prefer direct control.\n\n## Coding Agent Integration\n\nThis repository is a **single plugin marketplace** (`.claude-plugin/marketplace.json`) consumed by both **Claude Code** and **Grok** β same plugin id `cocoindex-code`, same `ccc` skill. Grok optionally activates the bundled hooks and MCP server with `--trust`; Claude Code users can install the same marketplace and rely on the skill alone or load hooks/MCP from the plugin as needed.\n\n### Skill (Recommended)\n\nInstall the `ccc` skill so your coding agent automatically uses semantic search when needed:\n\n```bash\nnpx skills add cocoindex-io/cocoindex-code\n```\n\nThat's it β no `ccc init` or `ccc index` needed. The skill teaches the agent to handle initialization, indexing, and searching on its own. It will automatically keep the index up to date as you work.\n\nThe agent uses semantic search automatically when it would be helpful. You can also nudge it explicitly β just ask it to search the codebase, e.g. *\"find how user sessions are managed\"*, or type `/ccc` to invoke the skill directly.\n\nWorks with [Claude Code](https://docs.anthropic.com/en/docs/claude-code) and other skill-compatible agents.\n\n#### Claude Code plugin marketplace\n\nFor Claude Code users, this repository is also a [plugin marketplace](https://code.claude.com/docs/en/plugin-marketplaces). Install the skill from inside Claude Code with:\n\n```text\n/plugin marketplace add cocoindex-io/cocoindex-code\n/plugin install cocoindex-code@cocoindex-code\n```\n\nThis bundles the same `ccc` skill, with version pinning and `/plugin marketplace update` for updates. The repository also ships `hooks/hooks.json` and `.mcp.json` for Grok (and Claude Code plugin installs that load those files); Claude users who want skill-only search can rely on the skill alone and add MCP manually in the [MCP Server](#mcp-server) section below instead of using the bundled `.mcp.json`.\n\n#### Grok plugin\n\nFor [Grok](https://github.com/xai-org/grok) users, install via Grok's plugin system. The plugin bundles three components:\n\n| Component | Purpose |\n|-----------|---------|\n| **Skill** (`skills/ccc/`) | Agent runs `ccc search` / `ccc index` via the CLI (same as Claude Code above) |\n| **Hook** (`hooks/hooks.json`) | `SessionStart` + `PostToolUse` (Edit/Write/β¦) β incremental `ccc index` when `.cocoindex_code/` exists |\n| **MCP** (`.mcp.json`) | `ccc mcp` stdio server β `search` tool with `refresh_index=true` by default |\n\nGrok does **not** import Claude's `enabledPlugins` or plugin cache; install separately even if you already use cocoindex in Claude Code.\n\n**Full install** (skill + hook + MCP):\n\n```bash\ngrok plugin marketplace add cocoindex-io/cocoindex-code\ngrok plugin install cocoindex-io/cocoindex-code --trust\ngrok plugin enable cocoindex-code\n```\n\nPrefer the GitHub shorthand (`cocoindex-io/cocoindex-code`) for install β `grok plugin install cocoindex-code` can fail when no marketplace plugin matches that bare name.\n\n`--trust` is required so Grok activates the plugin's hooks and MCP server (skills load when the plugin is enabled).\n\n**Skill-only** (match Claude Code β no auto-index hook, no MCP tool):\n\nInstall and enable as above, then disable the optional components:\n\n1. **Hooks** β open `/hooks`, select the `SessionStart` / `PostToolUse` hooks from `cocoindex-code`, press `Space` to disable.\n2. **MCP** β open `/mcps`, select `cocoindex-code`, press `Space` to disable; or persist in `~/.grok/config.toml`:\n\n```toml\n[mcp_servers.cocoindex-code]\nenabled = false\n```\n\nThe agent still owns indexing via the `ccc` skill (`ccc index` / `ccc search --refresh` when stale), same as Claude Code.\n\nTo avoid importing MCP servers from your Claude/Cursor user config (unrelated to this plugin):\n\n```toml\n[compat.claude]\nmcps = false\n\n[compat.cursor]\nmcps = false\n```\n\n#### Oh My Pi plugin\n\n[Oh My Pi](https://github.com/can1357/oh-my-pi) reads `.omp-plugin/marketplace.json` (preferred) and falls back to `.claude-plugin/marketplace.json`.\n\n```bash\nomp plugin marketplace add cocoindex-io/cocoindex-code\nomp plugin install cocoindex-code@cocoindex-code --scope project\n```\n\nThen restart the session (`/reload-plugins` does not reload extensions). Requires `ccc` on `PATH` (`uv tool install --upgrade 'cocoindex-code[full]'`).\n\n| Component | Purpose |\n|-----------|---------|\n| **Skill** (`skills/ccc/`) | Agent runs `ccc search` / `ccc index` via the CLI |\n| **Hook** (`extensions/ccc-index.ts` via `package.json#omp.extensions`) | `session_start` + post-edit `tool_result` β incremental `ccc index` when `.cocoindex_code/` exists |\n| **MCP** (`.mcp.json`) | `ccc mcp` stdio server |\n\nOMP does **not** execute Claude `hooks/hooks.json` command hooks. The TypeScript extension is the OMP equivalent of the Claude/Grok SessionStart + PostToolUse pair.\n\nSkill-only (no MCP, no auto-index): install the skill via `npx skills add cocoindex-io/cocoindex-code`, or run `ccc search` / `ccc index` from the shell.\n\n### MCP Server\n\nAlternatively, use `ccc mcp` to run as an MCP server:\n\n\nClaude Code\n\n```bash\nclaude mcp add cocoindex-code -- ccc mcp\n```\n\n\n\nCodex\n\n```bash\ncodex mcp add cocoindex-code -- ccc mcp\n```\n\n\n\nOpenCode\n\n```bash\nopencode mcp add\n```\nEnter MCP server name: `cocoindex-code`\nSelect MCP server type: `local`\nEnter command to run: `ccc mcp`\n\nOr use opencode.json:\n```json\n{\n \"$schema\": \"https://opencode.ai/config.json\",\n \"mcp\": {\n \"cocoindex-code\": {\n \"type\": \"local\",\n \"command\": [\n \"ccc\", \"mcp\"\n ]\n }\n }\n}\n```\n\n\n\nKilo Code\n\nAdd a local MCP server in `~/.config/kilo/kilo.jsonc`, `kilo.jsonc`, or `.kilo/kilo.jsonc`:\n\n```json\n{\n \"mcp\": {\n \"cocoindex-code\": {\n \"type\": \"local\",\n \"command\": [\"ccc\", \"mcp\"],\n \"enabled\": true\n }\n }\n}\n```\n\n\n\nOh My Pi\n\nPrefer the marketplace install above. Manual project MCP (`.omp/mcp.json`):\n\n```json\n{\n \"mcpServers\": {\n \"cocoindex-code\": {\n \"command\": \"ccc\",\n \"args\": [\"mcp\"]\n }\n }\n}\n```\n\n\nOnce configured, the agent automatically decides when semantic code search is helpful β finding code by description, exploring unfamiliar codebases, fuzzy/conceptual matches, or locating implementations without knowing exact names.\n\n> **Note:** The `cocoindex-code` command (without subcommand) still works as an MCP server for backward compatibility. It auto-creates settings from environment variables on first run.\n\n\nMCP Tool Reference\n\nWhen running as an MCP server (`ccc mcp`), the following tool is exposed:\n\n**`search`** β Search the codebase using semantic similarity.\n\n```\nsearch(\n query: str, # Natural language query or code snippet\n limit: int = 5, # Maximum results (1-100)\n offset: int = 0, # Pagination offset\n refresh_index: bool = True, # Refresh index before querying\n languages: list[str] | None = None, # Filter by language (e.g. [\"python\", \"typescript\"])\n paths: list[str] | None = None, # Filter by path glob (e.g. [\"src/utils/*\"])\n)\n```\n\nReturns matching code chunks with file path, language, code content, line numbers, and similarity score.\n\n\n## Manual CLI Usage\n\nYou can also use the CLI directly β useful for manual control, running indexing after changing settings, checking status, or searching outside an agent.\n\n```bash\nccc init # initialize project (creates settings)\nccc index # build the index\nccc search \"authentication logic\" # search!\n```\n\nThe background daemon starts automatically on first use.\n\n> **Tip:** You can skip `ccc init` and go straight to `ccc index` β it auto-initializes new projects with default settings. If global settings are missing too (first use on the machine), it walks you through the same model setup as `ccc init` when run interactively; non-interactive runs (scripts, hooks) still require a one-time `ccc init` first.\n\n### CLI Reference\n\n| Command | Description |\n|---------|-------------|\n| `ccc init` | Initialize a project β creates settings files, adds `.cocoindex_code/` to `.gitignore` |\n| `ccc index` | Build or update the index (auto-inits if needed). Shows streaming progress. |\n| `ccc search ` | Semantic search across the codebase |\n| `ccc grep [path]` | Structural code search by example (no index needed) |\n| `ccc status` | Show index stats (chunk count, file count, language breakdown) |\n| `ccc mcp` | Run as MCP server in stdio mode |\n| `ccc doctor` | Run diagnostics β checks settings, daemon, model, file matching, and index health |\n| `ccc reset` | Delete index databases. `--all` also removes settings. `-f` skips confirmation. |\n| `ccc version` | Print the CLI version |\n| `ccc daemon status` | Show daemon version, uptime, and loaded projects |\n| `ccc daemon restart` | Restart the background daemon |\n| `ccc daemon stop` | Stop the daemon |\n\n### Search Options\n\n```bash\nccc search database schema # basic search\nccc search --lang python --lang markdown schema # filter by language\nccc search --path 'src/utils/*' query handler # filter by path\nccc search --offset 10 --limit 5 database schema # pagination\nccc search --refresh database schema # update index first, then search\n```\n\nBy default, `ccc search` scopes results to your current working directory (relative to the project root). Use `--path` to override.\n\n### Structural Search (`ccc grep`)\n\n`ccc grep` finds code by **structure**, not text β you write a by-example pattern\nand it matches the syntax tree (via cocoindex's `code_match`), so formatting,\nwhitespace, and intervening tokens don't matter. It runs entirely locally: no\nindex, daemon, or embeddings required.\n\n```bash\nccc grep 'def \\NAME(\\(ARGS*\\)):' # every Python function def under the cwd\nccc grep 'foo(\\(ARGS*\\))' src/ # calls to foo(...) anywhere under src/\nccc grep 'fn \\NAME(\\(A*\\))' --lang rust # restrict to one language\nccc grep 'class \\NAME:' --path 'tests/**' # restrict to a path glob\nccc grep 'TODO(\\(A*\\))' path/to/file.py # a single file\n```\n\nMetavariables use the `\\` sigil: `\\NAME` captures one node, `\\(NAME*\\)` a run of\nsiblings, `\\_`/`\\*` match anonymously. The pattern is matched per language, so a\nsingle invocation scans every supported source file (others are skipped). Inside\nan initialized project, `ccc grep` honors the project's include/exclude patterns\nand `.gitignore`; otherwise it scans all supported source files under the path.\n\nResults stream to the terminal file-by-file as each match is found (in completion\norder, since files are matched in parallel) rather than all at once at the end.\nEach matching file shows its matched line range; under a TTY the path is colored,\nline numbers are dimmed, and the unmatched context around a match is dimmed so the\nmatch stands out.\n\n> **Note:** `ccc grep` relies on cocoindex's structural `code_match` feature.\n> Until it ships in a released cocoindex, run against a local cocoindex build.\n\n## Docker\n\nA Docker image is available for teams who want a reproducible, dependency-free\nsetup β no Python, `uv`, or system dependencies required on the host.\n\nThe recommended approach is a **persistent container**: start it once, and use\n`docker exec` to run CLI commands or connect MCP sessions to it. The daemon\ninside stays warm across sessions, so the embedding model is loaded only once.\n\n### Choosing an image\n\nTwo variants are published from each release:\n\n| Tag | Size | Embedding backends | When to pick |\n|---|---|---|---|\n| `cocoindex/cocoindex-code:latest` (slim, default) | ~450 MB | LiteLLM (cloud: OpenAI, Voyage, Gemini, Ollama, β¦) | Most users. Cloud-backed embeddings, smaller image, fast pulls. |\n| `cocoindex/cocoindex-code:full` | ~5 GB | sentence-transformers (local) + LiteLLM | When you want local embeddings without an API key, or an offline-ready container. Heavier because of torch + transformers. |\n\nThe rest of this section uses `:latest` β substitute `:full` in the `image:` /\n`docker run` commands if you want the full variant.\n\n> **Mac users running the `:full` variant:** local embedding inference is\n> CPU-only inside Docker, because Docker on macOS can't access Apple's Metal\n> (MPS) GPU. If you want local embeddings and fast inference, install\n> natively instead: `pipx install 'cocoindex-code[full]'`. The `:latest`\n> (slim) variant is unaffected β LiteLLM runs the model on the provider's\n> side, so Docker vs. native makes no difference.\n\n### Quick start β `docker compose up -d`\n\nBring it up in one line β no clone needed (bash / zsh):\n\n```bash\n# macOS / Windows\ndocker compose -f <(curl -L https://raw.githubusercontent.com/cocoindex-io/cocoindex-code/refs/heads/main/docker/docker-compose.yml) up -d\n\n# Linux (aligns file ownership on bind-mounted paths with your host user)\nPUID=$(id -u) PGID=$(id -g) docker compose -f <(curl -L https://raw.githubusercontent.com/cocoindex-io/cocoindex-code/refs/heads/main/docker/docker-compose.yml) up -d\n```\n\nOr grab [`docker/docker-compose.yml`](./docker/docker-compose.yml) and run `docker compose up -d` next to it (works on any shell, including Windows cmd / PowerShell).\n\nBy default your home directory is mounted into the container (set\n`COCOINDEX_HOST_WORKSPACE` to narrow this to a specific code folder). Index\ndata and the embedding model cache persist in a Docker volume across\nrestarts. Your global settings file at `$HOME/.cocoindex_code/global_settings.yml`\nis visible and editable on the host; edits take effect on your next `ccc` command.\n\n> **Pick a different image:** set `COCOINDEX_CODE_IMAGE` to override the\n> default. For example, the `:full` variant or GHCR:\n> ```bash\n> COCOINDEX_CODE_IMAGE=cocoindex/cocoindex-code:full docker compose up -d\n> COCOINDEX_CODE_IMAGE=ghcr.io/cocoindex-io/cocoindex-code:latest docker compose up -d\n> ```\n\n### Or: `docker run`\n\n\nDocker Desktop (macOS / Windows)\n\n```bash\ndocker run -d --name cocoindex-code \\\n --volume \"$HOME:/workspace\" \\\n --volume cocoindex-data:/var/cocoindex \\\n -e COCOINDEX_CODE_HOST_PATH_MAPPING=\"/workspace=$HOME\" \\\n cocoindex/cocoindex-code:latest\n```\n\n\n\nLinux (with PUID/PGID)\n\n```bash\ndocker run -d --name cocoindex-code \\\n -e PUID=$(id -u) -e PGID=$(id -g) \\\n --volume \"$HOME:/workspace\" \\\n --volume cocoindex-data:/var/cocoindex \\\n -e COCOINDEX_CODE_HOST_PATH_MAPPING=\"/workspace=$HOME\" \\\n cocoindex/cocoindex-code:latest\n```\n\n\n### Shell wrapper for `ccc` commands\n\nPaste this into `~/.bashrc` / `~/.zshrc` so `ccc` feels native on the host\nand picks up the right project based on your current directory:\n\n```bash\nccc() {\n docker exec -it -e COCOINDEX_CODE_HOST_CWD=\"$PWD\" cocoindex-code ccc \"$@\"\n}\n```\n\nNow `cd` into any project under your workspace and run `ccc init`, `ccc index`,\n`ccc search ...`, `ccc status`, etc. β it just works.\n\n### Connect your coding agent\n\n\nClaude Code\n\nRegister MCP from inside the target project so `$PWD` points there:\n\n```bash\nclaude mcp add cocoindex-code -- docker exec -i \\\n -e COCOINDEX_CODE_HOST_CWD=\"$PWD\" cocoindex-code ccc mcp\n```\n\nOr via `.mcp.json`:\n\n```json\n{\n \"mcpServers\": {\n \"cocoindex-code\": {\n \"type\": \"stdio\",\n \"command\": \"docker\",\n \"args\": [\n \"exec\",\n \"-i\",\n \"-e\",\n \"COCOINDEX_CODE_HOST_CWD=${PWD}\",\n \"cocoindex-code\",\n \"ccc\",\n \"mcp\"\n ]\n }\n }\n}\n```\n\n> Note: use `-i` (not `-it`). The `-t` flag allocates a terminal, which\n> interferes with MCP's JSON messaging over stdin/stdout β only add it for\n> interactive `ccc` commands like `ccc init`.\n\n\n\nCodex\n\n```bash\ncodex mcp add cocoindex-code -- docker exec -i \\\n -e COCOINDEX_CODE_HOST_CWD=\"$PWD\" cocoindex-code ccc mcp\n```\n\n\n### Upgrading from an older image\n\nEarlier images used separate `cocoindex-db` and `cocoindex-model-cache`\nvolumes; the current image consolidates them into a single `cocoindex-data`\nvolume. Before pulling the new image, drop the old container and volumes β\nindexes rebuild on your next `ccc index`, and the embedding model is\nre-populated automatically on first start:\n\n```bash\ndocker rm -f cocoindex-code\ndocker volume rm cocoindex-db cocoindex-model-cache\n```\n\n### Configuration via environment variables\n\nPass configuration to `docker run` / compose with `-e`:\n\n```bash\n# Extra extensions (e.g. Typesafe Config, SBT build files)\n-e COCOINDEX_CODE_EXTRA_EXTENSIONS=\"conf,sbt\"\n\n# Exclude build artefacts (Scala/SBT example)\n-e COCOINDEX_CODE_EXCLUDE_PATTERNS='[\"**/target/**\",\"**/.bloop/**\",\"**/.metals/**\"]'\n\n# Set an API key\n-e VOYAGE_API_KEY=your-key\n```\n\n> **Security note:** mounting `$HOME` gives the container read/write access\n> to everything under it. If that's too broad, bind-mount a narrower\n> directory instead (`COCOINDEX_HOST_WORKSPACE=/path/to/code`).\n\n### Build the image locally\n\n```bash\ndocker build -t cocoindex-code:local -f docker/Dockerfile .\n```\n\n## Features\n- **Semantic Code Search**: Find relevant code using natural language queries when grep doesn't work well, and save tokens immediately.\n- **Ultra Performant**: β‘ Built on top of ultra performant [Rust indexing engine](https://github.com/cocoindex-io/cocoindex). Only re-indexes changed files for fast updates.\n- **Multi-Language Support**: Python, JavaScript/TypeScript, Rust, Go, Java, C/C++, C#, SQL, Shell, and more.\n- **Embedded**: Portable and just works, no database setup required!\n- **Flexible Embeddings**: Local SentenceTransformers via the `[full]` extra (free, no API key!) or 100+ cloud providers via LiteLLM.\n\n## Configuration\n\nFor a detailed guide on choosing and configuring embedding models, see [EMBEDDINGS.md](EMBEDDINGS.md).\n\nConfiguration lives in two YAML files, both created automatically by `ccc init`.\n\n### User Settings (`~/.cocoindex_code/global_settings.yml`)\n\nShared across all projects. Controls the embedding model and environment variables for the daemon.\n\n```yaml\nembedding:\n provider: sentence-transformers # or \"litellm\"\n model: Snowflake/snowflake-arctic-embed-xs\n device: mps # optional: cpu, cuda, mps (auto-detected if omitted)\n min_interval_ms: 300 # optional: pace LiteLLM embedding requests to reduce 429s; defaults to 5 for LiteLLM\n mps_low_watermark_ratio: 0.4 # optional: PyTorch allocator soft limit\n mps_high_watermark_ratio: 0.5 # optional: PyTorch allocator hard limit\n\n # Optional extra kwargs passed to the embedder, separately for indexing vs query.\n # `ccc init` auto-populates these for known models (e.g. Cohere, Voyage, Nvidia NIM,\n # nomic-ai code-retrieval models, Snowflake arctic-embed).\n # indexing_params:\n # input_type: search_document # litellm: input_type\n # query_params:\n # input_type: search_query # sentence-transformers: prompt_name\n\nenvs: # extra environment variables for the daemon\n OPENAI_API_KEY: your-key # only needed if not already in your shell environment\n\ndaemon:\n idle_timeout_minutes: 180 # optional: exit the daemon after this long without client activity (default 180, 0 = never)\n keep_alive_with_mcp: true # optional: keep the daemon warm while an MCP client is connected (default true)\n```\n\n> **Note:** The daemon inherits your shell environment. If an API key (e.g. `OPENAI_API_KEY`) is already set as an environment variable, you don't need to duplicate it in `envs`. The `envs` field is only for values that aren't in your environment.\n\n> **Apple Silicon memory safety:** MPS SentenceTransformer calls use [CocoIndex's isolated GPU subprocess](https://github.com/cocoindex-io/cocoindex/blob/v1.0.18/python/cocoindex/_internal/runner.py), keeping the model loaded while isolating Metal allocations from the daemon. The low and high watermarks are ratios of PyTorch's recommended maximum working set; they default here to `0.4` and `0.5`. CocoIndex retries MPS out-of-memory failures with progressively smaller batches, and cocoindex-code [releases unused allocator cache](https://docs.pytorch.org/docs/stable/generated/torch.mps.empty_cache.html) after each index run. Explicit `COCOINDEX_RUN_GPU_IN_SUBPROCESS`, `PYTORCH_MPS_LOW_WATERMARK_RATIO`, and `PYTORCH_MPS_HIGH_WATERMARK_RATIO` environment variables take precedence over these defaults.\n\n> **Indexing concurrency:** Multiple projects may prepare indexes concurrently, while CocoIndex serializes their GPU calls through its single MPS subprocess. A search waits only when its own project still needs the initial index.\n\n> **Idle timeout:** the background daemon holds the embedding model in RAM, so it exits after `daemon.idle_timeout_minutes` without client activity and is restarted automatically on your next `ccc` command or MCP search. By default, a live MCP session sends periodic heartbeats so the daemon remains warm while your coding agent is connected. Set `daemon.keep_alive_with_mcp: false` to let the daemon idle-exit during long-lived MCP sessions and release the model between real requests. Set `idle_timeout_minutes: 0` to keep the daemon running forever.\n\n> **Custom location:** set `COCOINDEX_CODE_DIR` to place `global_settings.yml` somewhere other than `~/.cocoindex_code/` β useful if you want the file to live alongside your projects (e.g. on a synced folder).\n\n#### `indexing_params` / `query_params`\n\nSome embedding models expose different modes for documents vs queries (asymmetric retrieval). For example, Cohere's v3 models want `input_type: search_document` when embedding corpus content and `input_type: search_query` when embedding a user query; several SentenceTransformers models use `prompt_name: passage` / `prompt_name: query` for the same purpose. These knobs live under `indexing_params` and `query_params`:\n\n```yaml\nembedding:\n provider: litellm\n model: cohere/embed-english-v3.0\n indexing_params:\n input_type: search_document\n query_params:\n input_type: search_query\n```\n\n`ccc init` populates these automatically for models it recognizes β including all Cohere v3, Voyage, Nvidia NIM, Gemini embedding (`gemini/gemini-embedding-*`, `gemini/text-embedding-*`, `gemini/embedding-*` β LiteLLM auto-maps `input_type` to Gemini's `task_type`), `nomic-ai/CodeRankEmbed`, `nomic-ai/nomic-embed-code`, `nomic-ai/nomic-embed-text-v1`/`v1.5`, `mixedbread-ai/mxbai-embed-large-v1`, and the `Snowflake/snowflake-arctic-embed-*` family β and prints the chosen defaults. For other models, it leaves a commented-out template under `embedding:` so you can fill it in by hand.\n\nOpenAI embeddings (`text-embedding-3-*`, `text-embedding-ada-002`) are intentionally not in the list: they're symmetric and have no equivalent knob.\n\n**Accepted keys:** `prompt_name` (sentence-transformers) and `input_type` (litellm). Other keys are rejected at daemon startup with a clear error. Note: `dimensions` is intentionally not exposed here β output dimension must be identical for indexing and query, so it's a model-wide setting rather than a per-side knob.\n\n**Doctor checks both sides.** `ccc doctor` exercises the model once with `indexing_params` and once with `query_params`, reporting each as a separate `Model Check (indexing)` / `Model Check (query)` entry β so a misconfiguration on one side is diagnosable without hiding behind the other.\n\n**Legacy-bridge warning:** if you're upgrading from an earlier version and your `global_settings.yml` uses `nomic-ai/CodeRankEmbed` or `nomic-ai/nomic-embed-code` without `indexing_params` / `query_params`, the daemon continues to apply the previous behavior (`prompt_name: query` at query time) and prints a one-time warning asking you to make the setting explicit. You can silence the warning by adding an empty block such as `query_params: {}`.\n\n### Project Settings (`/.cocoindex_code/settings.yml`)\n\nPer-project. Controls which files to index.\n\n```yaml\ninclude_patterns:\n - \"**/*.py\"\n - \"**/*.js\"\n - \"**/*.ts\"\n - \"**/*.rs\"\n - \"**/*.go\"\n # ... (sensible defaults for 28+ file types)\n\nexclude_patterns:\n - \"**/.*\" # hidden directories\n - \"**/__pycache__\"\n - \"**/node_modules\"\n - \"**/dist\"\n # ...\n\nmax_file_size: 500KB # skip files larger than this (default: no limit)\n\nlanguage_overrides:\n - ext: inc # treat .inc files as PHP\n lang: php\n\nchunkers:\n - ext: toml # use a custom chunker for .toml files\n module: example_toml_chunker:toml_chunker\n```\n\n> `.cocoindex_code/` is automatically added to `.gitignore` during init.\n\n`max_file_size` keeps bundled or generated files out of the index without having\nto enumerate them in `exclude_patterns`. It accepts a plain byte count\n(`1048576`) or a size with a binary unit suffix: `B`, `KB`, `MB`, `GB`\n(case-insensitive, so `500KB` and `500 kb` are the same). The limit is\ninclusive, and omitting the key indexes files of any size. It applies wherever\nproject file matching does, so `ccc grep` skips the same files.\n\nAfter editing `include_patterns`, `exclude_patterns`, `max_file_size`, or `language_overrides`:\n\n- Run `ccc doctor` to preview which files match.\n- Run `ccc index` or `ccc search --refresh ...` to update the existing index.\n- You do not need to delete the index or restart the daemon for these file-matching changes.\n\nIf you add or change custom `chunkers`, restart the daemon first so the chunker registry is reloaded, then run `ccc index`.\n\nUse `chunkers` when you want to control how a file type is split into chunks before indexing.\n\n`module: example_toml_chunker:toml_chunker` means:\n- `example_toml_chunker` is a local Python module\n- `toml_chunker` is the function inside that module\n\nIn practice, this usually means:\n- you create a Python file in your project, for example `example_toml_chunker.py`\n- you add a function in that file\n- you point `settings.yml` at it with `module.path:function_name`\n\nThe function should use this signature:\n\n```python\nfrom pathlib import Path\nfrom cocoindex_code.chunking import Chunk\n\ndef my_chunker(path: Path, content: str) -> tuple[str | None, list[Chunk]]:\n ...\n```\n\n- `path` is the file being indexed\n- `content` is the full text of that file\n- return `language_override` as a string like `\"toml\"` if you want to override language detection\n- return `None` as `language_override` if you want to keep the detected language\n- return a `list[Chunk]` with the chunks you want stored in the index\n\nSee [`src/cocoindex_code/chunking.py`](./src/cocoindex_code/chunking.py) for the public types and [`tests/example_toml_chunker.py`](./tests/example_toml_chunker.py) for a complete example.\n\n## Embedding Models\n\nWith the `[full]` extra installed, `ccc init` defaults to a local SentenceTransformers model ([Snowflake/snowflake-arctic-embed-xs](https://huggingface.co/Snowflake/snowflake-arctic-embed-xs)) β no API key required. To use a different model, edit `~/.cocoindex_code/global_settings.yml`.\n\n> The `envs` entries below are only needed if the key isn't already in your shell environment β the daemon inherits your environment automatically.\n\n\nOllama (Local)\n\n```yaml\nembedding:\n model: ollama/nomic-embed-text\n```\n\nSet `OLLAMA_API_BASE` in `envs:` if your Ollama server is not at `http://localhost:11434`.\n\n\n\n\nOpenAI\n\n```yaml\nembedding:\n model: text-embedding-3-small\n min_interval_ms: 300 # optional: override the 5ms LiteLLM default\nenvs:\n OPENAI_API_KEY: your-api-key\n```\n\n\n\n\nOpenAI-compatible (custom endpoint)\n\nMany providers (vLLM, LM Studio, LocalAI, Together, Fireworks, DeepInfra, β¦) expose an OpenAI-compatible embedding API. Use the `openai/` prefix and point `OPENAI_BASE_URL` at your endpoint:\n\n```yaml\nembedding:\n model: openai/your-model-name\nenvs:\n OPENAI_BASE_URL: https://your-endpoint/v1\n OPENAI_API_KEY: your-api-key\n```\n\nDon't append `/embeddings` to the base URL β LiteLLM handles that.\n\n\n\n\nAzure OpenAI\n\n```yaml\nembedding:\n model: azure/your-deployment-name\nenvs:\n AZURE_API_KEY: your-api-key\n AZURE_API_BASE: https://your-resource.openai.azure.com\n AZURE_API_VERSION: \"2024-06-01\"\n```\n\n\n\n\nGemini\n\n```yaml\nembedding:\n model: gemini/gemini-embedding-001\nenvs:\n GEMINI_API_KEY: your-api-key\n```\n\n\n\n\nMistral\n\n```yaml\nembedding:\n model: mistral/mistral-embed\nenvs:\n MISTRAL_API_KEY: your-api-key\n```\n\n\n\n\nVoyage (Code-Optimized)\n\n```yaml\nembedding:\n model: voyage/voyage-code-3\nenvs:\n VOYAGE_API_KEY: your-api-key\n```\n\n\n\n\nCohere\n\n```yaml\nembedding:\n model: cohere/embed-v4.0\nenvs:\n COHERE_API_KEY: your-api-key\n```\n\n\n\n\nAWS Bedrock\n\n```yaml\nembedding:\n model: bedrock/amazon.titan-embed-text-v2:0\nenvs:\n AWS_ACCESS_KEY_ID: your-access-key\n AWS_SECRET_ACCESS_KEY: your-secret-key\n AWS_REGION_NAME: us-east-1\n```\n\n\n\n\nNebius\n\n```yaml\nembedding:\n model: nebius/BAAI/bge-en-icl\nenvs:\n NEBIUS_API_KEY: your-api-key\n```\n\n\n\nAny [LiteLLM-supported model](https://docs.litellm.ai/docs/embedding/supported_embedding) works. When using a LiteLLM model, set `provider: litellm` (or omit `provider` β LiteLLM is the default for non-`sentence-transformers` models). For the full list of env vars each provider reads (API keys, base URLs, regions, β¦), see LiteLLM's [Setting API Keys](https://docs.litellm.ai/docs/set_keys).\n\n### Local SentenceTransformers Models\n\nSet `provider: sentence-transformers` and use any [SentenceTransformers](https://www.sbert.net/) model (no API key required).\n\n**Example β general purpose text model:**\n```yaml\nembedding:\n provider: sentence-transformers\n model: nomic-ai/nomic-embed-text-v1.5\n```\n\n**GPU-optimised code retrieval:**\n\n[`nomic-ai/CodeRankEmbed`](https://huggingface.co/nomic-ai/CodeRankEmbed) delivers significantly better code retrieval than the default model. It is 137M parameters, requires ~1 GB VRAM, and has an 8192-token context window.\n\n```yaml\nembedding:\n provider: sentence-transformers\n model: nomic-ai/CodeRankEmbed\n```\n\n**Note:** Switching models requires re-indexing your codebase (`ccc reset && ccc index`) since the vector dimensions differ.\n\n## Supported Languages\n\n| Language | Aliases | File Extensions |\n|----------|---------|-----------------|\n| c | | `.c` |\n| cpp | c++ | `.cpp`, `.cc`, `.cxx`, `.h`, `.hpp` |\n| csharp | csharp, cs | `.cs` |\n| css | | `.css`, `.scss` |\n| dart | | `.dart` |\n| dtd | | `.dtd` |\n| elixir | | `.ex`, `.exs` |\n| fortran | f, f90, f95, f03 | `.f`, `.f90`, `.f95`, `.f03` |\n| go | golang | `.go` |\n| html | | `.html`, `.htm` |\n| java | | `.java` |\n| javascript | js | `.js` |\n| json | | `.json` |\n| kotlin | | `.kt`, `.kts` |\n| lua | | `.lua` |\n| markdown | md | `.md`, `.mdx` |\n| pascal | pas, dpr, delphi | `.pas`, `.dpr` |\n| php | | `.php` |\n| python | | `.py` |\n| r | | `.r` |\n| ruby | | `.rb` |\n| rust | rs | `.rs` |\n| scala | | `.scala` |\n| solidity | | `.sol` |\n| sql | | `.sql` |\n| svelte | | `.svelte` |\n| swift | | `.swift` |\n| toml | | `.toml` |\n| tsx | | `.tsx` |\n| typescript | ts | `.ts` |\n| vue | | `.vue` |\n| xml | | `.xml` |\n| yaml | | `.yaml`, `.yml` |\n\n### Custom Database Location\n\nBy default, index databases (`cocoindex.db` and `target_sqlite.db`) live alongside settings in `/.cocoindex_code/`. When running in Docker, you may want the databases on the container's native filesystem for performance (LMDB doesn't work well on mounted volumes) while keeping the source code and settings on a mounted volume.\n\nSet `COCOINDEX_CODE_DB_PATH_MAPPING` to remap database locations by path prefix:\n\n```bash\nCOCOINDEX_CODE_DB_PATH_MAPPING=/workspace=/db-files\n```\n\nWith this mapping, a project at `/workspace/myrepo` stores its databases in `/db-files/myrepo/` instead of `/workspace/myrepo/.cocoindex_code/`. Settings files remain in the original location.\n\nMultiple mappings are comma-separated and resolved in order (first match wins):\n\n```bash\nCOCOINDEX_CODE_DB_PATH_MAPPING=/workspace=/db-files,/workspace2=/db-files2\n```\n\nBoth source and target must be absolute paths. If no mapping matches, the default location is used.\n\n## Troubleshooting\n\nRun `ccc doctor` to diagnose common issues. It checks your settings, daemon health, embedding model, file matching, and index status β all in one command.\n\n### `sqlite3.Connection object has no attribute enable_load_extension`\n\nSome Python installations (e.g. the one pre-installed on macOS) ship with a SQLite library that doesn't enable extensions.\n\n**macOS fix:** Install Python through [Homebrew](https://brew.sh/):\n\n```bash\nbrew install python3\n```\n\nThen re-install cocoindex-code (see [Get Started](#get-started--zero-config-lets-go) for install options):\n\nUsing pipx:\n```bash\npipx install cocoindex-code # first install\npipx upgrade cocoindex-code # upgrade\n```\n\nUsing uv (install or upgrade):\n```bash\nuv tool install --upgrade cocoindex-code\n```\n\n### `MDB_MAP_FULL: Environment mapsize limit reached`\n\nThe index is stored in an LMDB database whose maximum size is fixed when the daemon starts. The default ceiling is **4 GiB**, which is plenty for most projects but can be exhausted by very large codebases (tens of thousands of files), especially with high-dimensional embedding models like `nomic-ai/CodeRankEmbed`.\n\nRaise the ceiling with the `COCOINDEX_LMDB_MAP_SIZE` environment variable (value in **bytes**). LMDB only grows the file as data is written, so a high limit doesn't pre-allocate disk β it's safe to set it generously:\n\n```yaml\n# ~/.cocoindex_code/global_settings.yml\nenvs:\n COCOINDEX_LMDB_MAP_SIZE: \"34359738368\" # 32 GiB (= 32 * 1024^3)\n```\n\nOr, if you prefer to set it in your shell environment (the daemon inherits it):\n\n```bash\nexport COCOINDEX_LMDB_MAP_SIZE=$((32 * 1024 * 1024 * 1024)) # 32 GiB\n```\n\nThe map size is read when the daemon starts, so restart it to pick up the change, then re-index:\n\n```bash\nccc daemon restart\nccc index\n```\n\n> This manual step is temporary. Once [cocoindex#2108](https://github.com/cocoindex-io/cocoindex/issues/2108) lands, the map size grows automatically when needed and `COCOINDEX_LMDB_MAP_SIZE` won't be necessary.\n\n## Legacy: Environment Variables\n\nIf you previously configured `cocoindex-code` via environment variables, the `cocoindex-code` MCP command still reads them and auto-migrates to YAML settings on first run. We recommend switching to the YAML settings for new setups.\n\n| Environment Variable | YAML Equivalent |\n|---------------------|-----------------|\n| `COCOINDEX_CODE_EMBEDDING_MODEL` | `embedding.model` in `global_settings.yml` |\n| `COCOINDEX_CODE_DEVICE` | `embedding.device` in `global_settings.yml` |\n| `COCOINDEX_CODE_ROOT_PATH` | Run `ccc init` in your project root instead |\n| `COCOINDEX_CODE_EXCLUDED_PATTERNS` | `exclude_patterns` in project `settings.yml` |\n| `COCOINDEX_CODE_EXTRA_EXTENSIONS` | `include_patterns` + `language_overrides` in project `settings.yml` |\n\n## Telemetry\n\n`cocoindex-code` sends anonymous usage telemetry through CocoIndex so we can see how the tool is used in aggregate and prioritize improvements. The events identify themselves as `application: cocoindex-code`.\n\nWe **do not** collect your source code, file paths, queries, search results, embeddings, settings, or any other content from your codebase or environment.\n\nTo opt out, set:\n\n```bash\nexport COCOINDEX_DISABLE_USAGE_TRACKING=1\n```\n\n## Large codebase / Enterprise\n[CocoIndex](https://github.com/cocoindex-io/cocoindex) is an ultra efficient indexing engine that also works on large codebases at scale for enterprises. In enterprise scenarios it is a lot more efficient to share indexes with teammates when there are large or many repos. We also have advanced features like branch dedupe etc designed for enterprise users.\n\n> Indexing a very large codebase and hitting `MDB_MAP_FULL`? Raise the LMDB map size β see [`MDB_MAP_FULL: Environment mapsize limit reached`](#mdb_map_full-environment-mapsize-limit-reached) under Troubleshooting.\n\nIf you need help with remote setup, please email our maintainer linghua@cocoindex.io, happy to help!\n\n## Contributing\n\nWe welcome contributions! This project uses [uv](https://docs.astral.sh/uv/getting-started/installation/) for development, and every PR is gated on the same lint, format, type-check, and test suite in CI. **Please run these checks locally before opening a PR** β failing pre-commit checks are the most common cause of red CI on incoming PRs.\n\n### 1. Install the dev dependencies\n\nAfter installing [uv](https://docs.astral.sh/uv/getting-started/installation/), sync the project. This installs everything the checks need β including [prek](https://github.com/j178/prek), a fast pre-commit runner, plus Ruff, mypy, and pytest:\n\n```bash\nuv sync\n```\n\n### 2. Run all checks before every PR\n\nRun the full hook suite across all files β this is exactly what CI runs:\n\n```bash\nuv run prek run --all-files\n```\n\nIt runs trailing-whitespace/end-of-file fixes, Ruff lint (`--fix`) and format, `uv.lock` validation, mypy type checking, and the pytest suite. Fix anything it reports (Ruff auto-fixes most lint/format issues for you), re-run until it passes, then push.\n\n### 3. (Optional) Run automatically on each commit\n\nTo have the same checks run on every `git commit`, install the git hook once:\n\n```bash\nuv run prek install\n```\n\nFor more details, see our [contributing guide](https://cocoindex.io/docs/contributing/guide).\n\n## License\n\nApache-2.0\n",
"bytes": 41851,
"sha": "6dbd4e9d33a5ccac4894d59a7124ea273a9de2fadccefdf935cd4c884fd3e070",
"repo_slug": "cocoindex-io/cocoindex-code",
"fonte": "repo",
"truncated": false,
"api": "https://agentalog.com/api/listings/skl_cocoindex_io_cocoindex_code_ccc_78976be2/readme"
}