​

Key Concepts

​

Source Types

​

S3

​

GitHub Repositories

​

Direct Sync

​

Platform Sync

​

Web

​

Sources

{
  "database_name": "string",
  "embedding": {
    "dense": {
        "model": "Qwen/Qwen3-Embedding-0.6B"
    }
  }
}

• database_name defines the Chroma database in which collections should be created by invocations run on this source. A database must exist before creating sources that point to it.
• embedding.dense.model defines the embedding model that should be used to generate dense embeddings for chunked documents. Currently, only the Qwen3-Embedding-0.6B model is supported, but if there is a model you would like to use, please let us know by reaching out to engineering@trychroma.com.

{
  "embedding": {
    "dense": {
      "model": "Qwen/Qwen3-Embedding-0.6B"
    },
    "sparse": {
      "model": "Chroma/BM25",
      "key": "sparse_embedding"
    }
  }
}

• embedding.sparse.model defines the sparse embedding model. Supported models: Chroma/BM25, prithivida/Splade_PP_en_v1.
• embedding.sparse.key defines the metadata key under which sparse embeddings are stored.

{
  "chunking": {
    "type": "tree_sitter",
    "max_size_bytes": 8192
  }
}

• chunking.type can be tree_sitter (syntax-aware, with max_size_bytes) or lines (line-based, with max_lines and max_size_bytes).

​

S3

{
    "bucket_name": "string",
    "region": "string",
    "collection_name": "string",
    "aws_credential_id": 0,
    "path_prefix": "string",
    "auto_sync": "none"
}

• bucket_name is the name of the S3 bucket to sync from.
• region is the AWS region of the bucket.
• collection_name is the default target collection name for synced data.
• aws_credential_id is the ID of AWS credentials.
• path_prefix (optional) limits which S3 keys can be synced. Only keys starting with this prefix are allowed.
• auto_sync (optional) sets the auto-sync mode: none (default), direct, or metadata. See S3 Auto-Sync.

​

GitHub Repositories

{
	"repository": "string",
	"app_id": "string" | null, // optional
	"include_globs": ["string", ...] | null, // optional
}

• repository defines the GitHub repository whose code should be synced. This must be the forward slash-separated combination of the repository owner’s GitHub username and the repository name (e.g., chroma-core/chroma). Note that changing a repository name after creating a Chroma Sync source for it will result in invocations on that source failing, so a new source with the updated repository name must be created.
• app_id defines the GitHub App ID of the GitHub App that has access to the provided repository. This parameter should only be supplied if the provided repository is private.
• include_globs defines a set of glob patterns for which matching files should be synced. If this parameter is not provided, files matching "*" will be synced. Note that Chroma will not sync binary data, images, and other large or non-UTF-8 files.

​

Web

{
    "starting_url": "https://docs.trychroma.com",
    // all below are optional
    "page_limit": 5,
    "include_path_regexes": ["/cloud/*"],
    "exclude_path_regexes": ["/blog/*"],
    "max_depth": 2
}

​

Invocations

{
	"target_collection_name": "string"
}

• target_collection_name defines the name of the Chroma collection in which synced data should be stored. This is required for GitHub and Web sources. For S3 sources, it is optional and defaults to the collection_name configured on the source. The target must be a collection that does not already exist with synced data. Chroma Sync uses the metadata key finished_ingest to indicate whether a collection contains synced data. If an invocation creation request is received for a collection with metadata in which this key is present and set to true, the API will return a 409 Conflict.

​

S3

{
    "object_key": "string",
    "custom_id": "string",
    "metadata": {},
    "target_collection_name": "string"
}

• object_key (required) is the full S3 object key to sync. Must include the path_prefix if one is configured on the source.
• custom_id (optional) is a custom document ID (max 120 bytes). Chunk IDs become custom_id-{chunk} instead of sha256(object_key)-{chunk}.
• metadata (optional) is additional metadata merged with standard chunk metadata. Values must be scalars (string, number, boolean, or null).
• target_collection_name (optional) overrides the source’s collection_name. If not provided, defaults to the collection_name configured on the source.

​

GitHub Repositories

{
	"ref_identifier": {
		"$oneOf": {
			"branch": "string",
			"sha": "string"
		}
	}
}

• ref_identifier is either the commit SHA-256 or the name of the branch from which to retrieve the code to be synced. If a branch is provided, the code will be retrieved from the branch’s latest commit.