AST-based semantic code search that just works
A lightweight, effective (AST-based) semantic code search tool for your codebase. Built on CocoIndex — a Rust-based ultra performant data transformation engine. Use it from the CLI, or integrate with Claude, Codex, Cursor — any coding agent — via Skill or MCP.
- Instant token saving by 70%.
- 1 min setup — install and go, zero config needed!
🌟 Please help star CocoIndex if you like this project!
Deutsch |
English |
Español |
français |
日本語 |
한국어 |
Português |
Русский |
中文
Get Started — zero config, let's go!
Install
Using pipx:```bash
pipx install 'cocoindex-code[full]' # batteries included (local embeddings)
pipx upgrade cocoindex-code # upgrade
Using uv:```bash
uv tool install --upgrade 'cocoindex-code[full]'
Two install styles — they mirror the Docker image variants of the same names:
- cocoindex-code[full] — 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.
- cocoindex-code (slim) — 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).
Next, set up your coding agent integration — or jump to Manual CLI Usage if you prefer direct control.
Coding Agent Integration
This repository is a single plugin marketplace (.claude-plugin/marketplace.json) consumed by both Claude Code and Grok — 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.
Skill (Recommended)
Install the ccc skill so your coding agent automatically uses semantic search when needed:```bash
npx skills add cocoindex-io/cocoindex-code
That's it — 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.
The agent uses semantic search automatically when it would be helpful. You can also nudge it explicitly — just ask it to search the codebase, e.g. *"find how user sessions are managed"*, or type `/ccc` to invoke the skill directly.
Works with Claude Code and other skill-compatible agents.
#### Claude Code plugin marketplace
For Claude Code users, this repository is also a plugin marketplace. Install the skill from inside Claude Code with:```text
/plugin marketplace add cocoindex-io/cocoindex-code
/plugin install cocoindex-code@cocoindex-code
This 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 section below instead of using the bundled .mcp.json.
Grok plugin
For Grok users, install via Grok's plugin system. The plugin bundles three components:
| Component | Purpose |
|---|---|
Skill (skills/ccc/) |
Agent runs ccc search / ccc index via the CLI (same as Claude Code above) |
Hook (hooks/hooks.json) |
SessionStart + PostToolUse (Edit/Write/…) → incremental ccc index when .cocoindex_code/settings.yml exists (never in the directory holding ccc's user settings, ~ by default) |
MCP (.mcp.json) |
ccc mcp stdio server — search tool with refresh_index=true by default |
Grok does not import Claude's enabledPlugins or plugin cache; install separately even if you already use cocoindex in Claude Code.
Full install (skill + hook + MCP):```bash
grok plugin marketplace add cocoindex-io/cocoindex-code
grok plugin install cocoindex-io/cocoindex-code --trust
grok plugin enable cocoindex-code
Prefer the GitHub shorthand (`cocoindex-io/cocoindex-code`) for install — `grok plugin install cocoindex-code` can fail when no marketplace plugin matches that bare name.
`--trust` is required so Grok activates the plugin's hooks and MCP server (skills load when the plugin is enabled).
**Skill-only** (match Claude Code — no auto-index hook, no MCP tool):
Install and enable as above, then disable the optional components:
1. **Hooks** — open `/hooks`, select the `SessionStart` / `PostToolUse` hooks from `cocoindex-code`, press `Space` to disable.
2. **MCP** — open `/mcps`, select `cocoindex-code`, press `Space` to disable; or persist in `~/.grok/config.toml`:```toml
[mcp_servers.cocoindex-code]
enabled = false
The agent still owns indexing via the ccc skill (ccc index / ccc search --refresh when stale), same as Claude Code.
To avoid importing MCP servers from your Claude/Cursor user config (unrelated to this plugin):```toml
[compat.claude]
mcps = false
[compat.cursor]
mcps = false
#### Oh My Pi plugin
Oh My Pi reads `.omp-plugin/marketplace.json` (preferred) and falls back to `.claude-plugin/marketplace.json`.```bash
omp plugin marketplace add cocoindex-io/cocoindex-code
omp plugin install cocoindex-code@cocoindex-code --scope project
Then restart the session (/reload-plugins does not reload extensions). Requires ccc on PATH (uv tool install --upgrade 'cocoindex-code[full]').
| Component | Purpose |
|---|---|
Skill (skills/ccc/) |
Agent runs ccc search / ccc index via the CLI |
Hook (extensions/ccc-index.ts via package.json#omp.extensions) |
session_start + post-edit tool_result → incremental ccc index when .cocoindex_code/settings.yml exists (never in the directory holding ccc's user settings, ~ by default) |
MCP (.mcp.json) |
ccc mcp stdio server |
OMP does not execute Claude hooks/hooks.json command hooks. The TypeScript extension is the OMP equivalent of the Claude/Grok SessionStart + PostToolUse pair.
Skill-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.
MCP Server
Alternatively, use ccc mcp to run as an MCP server:
Claude Code```bash
claude mcp add cocoindex-code -- ccc mcp
Codex```bash
codex mcp add cocoindex-code -- ccc mcp
OpenCode```bash
opencode mcp add
Enter MCP server name: `cocoindex-code`
Select MCP server type: `local`
Enter command to run: `ccc mcp`
Or use opencode.json:```json
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"cocoindex-code": {
"type": "local",
"command": [
"ccc", "mcp"
]
}
}
}
Kilo Code
Add a local MCP server in ~/.config/kilo/kilo.jsonc, kilo.jsonc, or .kilo/kilo.jsonc:```json
{
"mcp": {
"cocoindex-code": {
"type": "local",
"command": ["ccc", "mcp"],
"enabled": true
}
}
}
Oh My Pi
Prefer the marketplace install above. Manual project MCP (`.omp/mcp.json`):```json
{
"mcpServers": {
"cocoindex-code": {
"command": "ccc",
"args": ["mcp"]
}
}
}
Once configured, the agent automatically decides when semantic code search is helpful — finding code by description, exploring unfamiliar codebases, fuzzy/conceptual matches, or locating implementations without knowing exact names.
> 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.
MCP Tool Reference
When running as an MCP server (ccc mcp), the following tool is exposed:
search — Search the codebase using semantic similarity.```
search(
query: str, # Natural language query or code snippet
limit: int = 5, # Maximum results (1-100)
offset: int = 0, # Pagination offset
refresh_index: bool = True, # Refresh index before querying
languages: list[str] | None = None, # Filter by language (e.g. ["python", "typescript"])
paths: list[str] | None = None, # Filter by path glob (e.g. ["src/utils/*"])
)
Returns matching code chunks with file path, language, code content, line numbers, and similarity score.
## Manual CLI Usage
You can also use the CLI directly — useful for manual control, running indexing after changing settings, checking status, or searching outside an agent.```bash
ccc init # initialize project (creates settings)
ccc index # build the index
ccc search "authentication logic" # search!
The background daemon starts automatically on first use.
> Tip: You can skip ccc init and go straight to ccc index — 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.
> Note:
ccc index never auto-initializes the directory that holds ccc's user settings directory: your home directory by default (~/.cocoindex_code holds the global settings and daemon files), or the parent of $COCOINDEX_CODE_DIR when that directory is named .cocoindex_code. A project there would cover every repo below it. The editor hooks skip that directory even if it is already a project. If an older ccc made your home directory a project (~/.cocoindex_code/settings.yml exists), run ccc reset --all in your home directory: it removes that project's settings and indexes and keeps your global settings. To index your whole home directory on purpose, run ccc init there.
CLI Reference
| Command | Description |
|---|---|
ccc init |
Initialize a project — creates settings files, adds .cocoindex_code/ to .gitignore |
ccc index |
Build or update the index (auto-inits if needed). Shows streaming progress. |
ccc search |
Semantic search across the codebase |
ccc grep [path] |
Structural code search by example (no index needed) |
ccc status |
Show index stats (chunk count, file count, language breakdown) |
ccc mcp |
Run as MCP server in stdio mode |
ccc doctor |
Run diagnostics — checks settings, daemon, model, file matching, and index health |
ccc reset |
Delete index databases. --all also removes settings. -f skips confirmation. |
ccc version |
Print the CLI version |
ccc daemon status |
Show daemon version, uptime, and loaded projects |
ccc daemon restart |
Restart the background daemon |
ccc daemon stop |
Stop the daemon |
Search Options```bash
ccc search database schema # basic search
ccc search --lang python --lang markdown schema # filter by language
ccc search --path 'src/utils/*' query handler # filter by path
ccc search --offset 10 --limit 5 database schema # pagination
ccc search --refresh database schema # update index first, then search
By default, `ccc search` scopes results to your current working directory (relative to the project root). Use `--path` to override.
### Structural Search (`ccc grep`)
`ccc grep` finds code by **structure**, not text — you write a by-example pattern
and it matches the syntax tree (via cocoindex's `code_match`), so formatting,
whitespace, and intervening tokens don't matter. It runs entirely locally: no
index, daemon, or embeddings required.```bash
ccc grep 'def \NAME(\(ARGS*\)):' # every Python function def under the cwd
ccc grep 'foo(\(ARGS*\))' src/ # calls to foo(...) anywhere under src/
ccc grep 'fn \NAME(\(A*\))' --lang rust # restrict to one language
ccc grep 'class \NAME:' --path 'tests/**' # restrict to a path glob
ccc grep 'TODO(\(A*\))' path/to/file.py # a single file
Metavariables use the \ sigil: \NAME captures one node, \(NAME*\) a run of
siblings, \_/\* match anonymously. The pattern is matched per language, so a
single invocation scans every supported source file (others are skipped). Inside
an initialized project, ccc grep honors the project's include/exclude patterns
and .gitignore; otherwise it scans all supported source files under the path.
Results stream to the terminal file-by-file as each match is found (in completion
order, since files are matched in parallel) rather than all at once at the end.
Each matching file shows its matched line range; under a TTY the path is colored,
line numbers are dimmed, and the unmatched context around a match is dimmed so the
match stands out.
> Note:
ccc grep relies on cocoindex's structural code_match feature.
> Until it ships in a released cocoindex, run against a local cocoindex build.
Docker
A Docker image is available for teams who want a reproducible, dependency-free
setup — no Python, uv, or system dependencies required on the host.
The recommended approach is a persistent container: start it once, and use
docker exec to run CLI commands or connect MCP sessions to it. The daemon
inside stays warm across sessions, so the embedding model is loaded only once.
Choosing an image
Two variants are published from each release:
| Tag | Size | Embedding backends | When to pick |
|---|---|---|---|
cocoindex/cocoindex-code:latest (slim, default) |
~450 MB | LiteLLM (cloud: OpenAI, Voyage, Gemini, Ollama, …) | Most users. Cloud-backed embeddings, smaller image, fast pulls. |
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. |
The rest of this section uses :latest — substitute :full in the image: /
docker run commands if you want the full variant.
> Mac users running the :full variant: local embedding inference is
> CPU-only inside Docker, because Docker on macOS can't access Apple's Metal
> (MPS) GPU. If you want local embeddings and fast inference, install
> natively instead: pipx install 'cocoindex-code[full]'. The :latest
> (slim) variant is unaffected — LiteLLM runs the model on the provider's
> side, so Docker vs. native makes no difference.
Quick start — docker compose up -d
Bring it up in one line — no clone needed (bash / zsh):```bash
macOS / Windows
docker compose -f <(curl -L https://raw.githubusercontent.com/cocoindex-io/cocoindex-code/refs/heads/main/docker/docker-compose.yml) up -d
Linux (aligns file ownership on bind-mounted paths with your host user)
PUID=$(id -u) PGID=$(id -g) docker compose -f <(curl -L https://raw.githubusercontent.com/cocoindex-io/cocoindex-code/refs/heads/main/docker/docker-compose.yml) up -d
Or grab `docker/docker-compose.yml` and run `docker compose up -d` next to it (works on any shell, including Windows cmd / PowerShell).
By default your home directory is mounted into the container (set
`COCOINDEX_HOST_WORKSPACE` to narrow this to a specific code folder). Index
data and the embedding model cache persist in a Docker volume across
restarts. Your global settings file at `$HOME/.cocoindex_code/global_settings.yml`
is visible and editable on the host; edits take effect on your next `ccc` command.
> **Pick a different image:** set `COCOINDEX_CODE_IMAGE` to override the
> default. For example, the `:full` variant or GHCR:
>```bash
> COCOINDEX_CODE_IMAGE=cocoindex/cocoindex-code:full docker compose up -d
> COCOINDEX_CODE_IMAGE=ghcr.io/cocoindex-io/cocoindex-code:latest docker compose up -d
> ```
### Or: `docker run`
Docker Desktop (macOS / Windows)```bash
docker run -d --name cocoindex-code \
--volume "$HOME:/workspace" \
--volume cocoindex-data:/var/cocoindex \
-e COCOINDEX_CODE_HOST_PATH_MAPPING="/workspace=$HOME" \
cocoindex/cocoindex-code:latest
Linux (with PUID/PGID)```bash
docker run -d --name cocoindex-code \
-e PUID=$(id -u) -e PGID=$(id -g) \
--volume "$HOME:/workspace" \
--volume cocoindex-data:/var/cocoindex \
-e COCOINDEX_CODE_HOST_PATH_MAPPING="/workspace=$HOME" \
cocoindex/cocoindex-code:latest
### Shell wrapper for `ccc` commands
Paste this into `~/.bashrc` / `~/.zshrc` so `ccc` feels native on the host
and picks up the right project based on your current directory:```bash
ccc() {
docker exec -it -e COCOINDEX_CODE_HOST_CWD="$PWD" cocoindex-code ccc "$@"
}
Now cd into any project under your workspace and run ccc init, ccc index,
ccc search ..., ccc status, etc. — it just works.
Connect your coding agent
Claude Code
Register MCP from inside the target project so $PWD points there:```bash
claude mcp add cocoindex-code -- docker exec -i \
-e COCOINDEX_CODE_HOST_CWD="$PWD" cocoindex-code ccc mcp
Or via `.mcp.json`:```json
{
"mcpServers": {
"cocoindex-code": {
"type": "stdio",
"command": "docker",
"args": [
"exec",
"-i",
"-e",
"COCOINDEX_CODE_HOST_CWD=${PWD}",
"cocoindex-code",
"ccc",
"mcp"
]
}
}
}
> Note: use -i (not -it). The -t flag allocates a terminal, which
> interferes with MCP's JSON messaging over stdin/stdout — only add it for
> interactive ccc commands like ccc init.
Codex```bash
codex mcp add cocoindex-code -- docker exec -i \
-e COCOINDEX_CODE_HOST_CWD="$PWD" cocoindex-code ccc mcp
### Upgrading from an older image
Earlier images used separate `cocoindex-db` and `cocoindex-model-cache`
volumes; the current image consolidates them into a single `cocoindex-data`
volume. Before pulling the new image, drop the old container and volumes —
indexes rebuild on your next `ccc index`, and the embedding model is
re-populated automatically on first start:```bash
docker rm -f cocoindex-code
docker volume rm cocoindex-db cocoindex-model-cache
Configuration via environment variables
Pass configuration to docker run / compose with -e:```bash
Extra extensions (e.g. Typesafe Config, SBT build files)
-e COCOINDEX_CODE_EXTRA_EXTENSIONS="conf,sbt"
Exclude build artefacts (Scala/SBT example)
-e COCOINDEX_CODE_EXCLUDE_PATTERNS='["/target/","/.bloop/","/.metals/"]'
Set an API key
-e VOYAGE_API_KEY=your-key
> **Security note:** mounting `$HOME` gives the container read/write access
> to everything under it. If that's too broad, bind-mount a narrower
> directory instead (`COCOINDEX_HOST_WORKSPACE=/path/to/code`).
### Build the image locally```bash
docker build -t cocoindex-code:local -f docker/Dockerfile .
Features
- Semantic Code Search: Find relevant code using natural language queries when grep doesn't work well, and save tokens immediately.
- Ultra Performant: ⚡ Built on top of ultra performant Rust indexing engine. Only re-indexes changed files for fast updates.
- Embedded: Portable and just works, no database setup required!
- Flexible Embeddings: Local SentenceTransformers via the
[full]extra (free, no API key!) or 100+ cloud providers via LiteLLM.
Configuration
For a detailed guide on choosing and configuring embedding models, see EMBEDDINGS.md.
Configuration lives in two YAML files, both created automatically by ccc init.
User Settings (~/.cocoindex_code/global_settings.yml)
Shared across all projects. Controls the embedding model and environment variables for the daemon.```yaml
embedding:
provider: sentence-transformers # or "litellm"
model: Snowflake/snowflake-arctic-embed-xs
device: mps # optional: cpu, cuda, mps (auto-detected if omitted)
min_interval_ms: 300 # optional: pace LiteLLM embedding requests to reduce 429s; defaults to 5 for LiteLLM
mps_low_watermark_ratio: 0.4 # optional: PyTorch allocator soft limit
mps_high_watermark_ratio: 0.5 # optional: PyTorch allocator hard limit
# Optional extra kwargs passed to the embedder, separately for indexing vs query.
# ccc init auto-populates these for known models (e.g. Cohere, Voyage, Nvidia NIM,
# nomic-ai code-retrieval models, Snowflake arctic-embed).
# indexing_params:
# input_type: search_document # litellm: input_type
# query_params:
# input_type: search_query # sentence-transformers: prompt_name
envs: # extra environment variables for the daemon
OPENAI_API_KEY: your-key # only needed if not already in your shell environment
daemon:
idle_timeout_minutes: 180 # optional: exit the daemon after this long without client activity (default 180, 0 = never)
keep_alive_with_mcp: true # optional: keep the daemon warm while an MCP client is connected (default true)
> **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.
> **Apple Silicon memory safety:** MPS SentenceTransformer calls use CocoIndex's isolated GPU subprocess, 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 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.
> **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.
> **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.
> **Custom location:** set `COCOINDEX_CODE_DIR` to place `global_settings.yml` somewhere other than `~/.cocoindex_code/` — useful if you want the file to live alongside your projects (e.g. on a synced folder).
#### `indexing_params` / `query_params`
Some 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`:```yaml
embedding:
provider: litellm
model: cohere/embed-english-v3.0
indexing_params:
input_type: search_document
query_params:
input_type: search_query
ccc init populates these automatically for models it recognizes — including all Cohere v3, Voyage, Nvidia NIM, Gemini embedding (gemini/gemini-embedding-*, gemini/text-embedding-*, gemini/embedding-* — 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 — and prints the chosen defaults. For other models, it leaves a commented-out template under embedding: so you can fill it in by hand.
OpenAI embeddings (text-embedding-3-*, text-embedding-ada-002) are intentionally not in the list: they're symmetric and have no equivalent knob.
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 — output dimension must be identical for indexing and query, so it's a model-wide setting rather than a per-side knob.
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 — so a misconfiguration on one side is diagnosable without hiding behind the other.
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: {}.
Project Settings (/.cocoindex_code/settings.yml)
Per-project. Controls which files to index.```yaml
include_patterns:
- "/*.py"
- "/.js"
- "/.ts"
- "/*.rs"
- "/*.go"
# ... (sensible defaults for 28+ file types)
exclude_patterns:
- "/.*" # hidden directories
- "/pycache"
- "/node_modules"
- "/dist"
# ...
max_file_size: 500KB # skip files larger than this (default: no limit)
language_overrides:
- ext: inc # treat .inc files as PHP
lang: php
chunkers:
- ext: toml # use a custom chunker for .toml files
module: example_toml_chunker:toml_chunker
> `.cocoindex_code/` is automatically added to `.gitignore` during init.
`max_file_size` keeps bundled or generated files out of the index without having
to enumerate them in `exclude_patterns`. It accepts a plain byte count
(`1048576`) or a size with a binary unit suffix: `B`, `KB`, `MB`, `GB`
(case-insensitive, so `500KB` and `500 kb` are the same). The limit is
inclusive, and omitting the key indexes files of any size. It applies wherever
project file matching does, so `ccc grep` skips the same files.
After editing `include_patterns`, `exclude_patterns`, `max_file_size`, or `language_overrides`:
- Run `ccc doctor` to preview which files match.
- Run `ccc index` or `ccc search --refresh ...` to update the existing index.
- You do not need to delete the index or restart the daemon for these file-matching changes.
- After any `language_overrides` change, the next run re-chunks all files, including files whose content did not change; the run reports them as reprocessed. Chunks whose text did not change keep their cached embeddings.
If you add or change custom `chunkers`, or edit a chunker module, restart the daemon (`ccc daemon restart`), then run `ccc index`. The daemon imports chunkers once at startup, so until it restarts it keeps running the old chunker code. The first run after the restart re-chunks all files, not only the ones the changed entry applies to; chunks whose text did not change keep their cached embeddings. A chunker counts as changed when its `module:` value changes or when the file of the module it names changes. Edits to other modules that file imports are not detected: after such an edit, also change the chunker's own module file (for example, bump a version comment in it).
Use `chunkers` when you want to control how a file type is split into chunks before indexing.
`module: example_toml_chunker:toml_chunker` means:
- `example_toml_chunker` is a local Python module
- `toml_chunker` is the function inside that module
In practice, this usually means:
- you create a Python file in your project, for example `example_toml_chunker.py`
- you add a function in that file
- you point `settings.yml` at it with `module.path:function_name`
The function should use this signature:```python
from pathlib import Path
from cocoindex_code.chunking import Chunk
def my_chunker(path: Path, content: str) -> tuple[str | None, list[Chunk]]:
...
pathis the file being indexedcontentis the full text of that file- return
language_overrideas a string like"toml"if you want to override language detection - return
Noneaslanguage_overrideif you want to keep the detected language - return a
list[Chunk]with the chunks you want stored in the index
See src/cocoindex_code/chunking.py for the public types and tests/example_toml_chunker.py for a complete example.
Embedding Models
With the [full] extra installed, ccc init defaults to a local SentenceTransformers model (Snowflake/snowflake-arctic-embed-xs) — no API key required. To use a different model, edit ~/.cocoindex_code/global_settings.yml.
> The envs entries below are only needed if the key isn't already in your shell environment — the daemon inherits your environment automatically.
Ollama (Local)```yaml
embedding:
model: ollama/nomic-embed-text
Set `OLLAMA_API_BASE` in `envs:` if your Ollama server is not at `http://localhost:11434`.
OpenAI```yaml
embedding:
model: text-embedding-3-small
min_interval_ms: 300 # optional: override the 5ms LiteLLM default
envs:
OPENAI_API_KEY: your-api-key
OpenAI-compatible (custom endpoint)
Many providers (vLLM, LM Studio, LocalAI, Together, Fireworks, DeepInfra, …) expose an OpenAI-compatible embedding API. Use the openai/ prefix and point OPENAI_BASE_URL at your endpoint:```yaml
embedding:
model: openai/your-model-name
envs:
…(正文过长,页面只展示前面部分,完整内容见仓库)
```
