Code Index SDK

Source Code: src/gaia/code_index/sdk.py, src/gaia/code_index/parsers.py, and src/gaia/agents/tools/code_index_tools.py

Looking for usage docs? See the Code Index guide for the flagship-agent skill walkthrough and architecture overview.

CodeIndexSDK indexes repository files (no git history, no PRs) using Lemonade Server embeddings and FAISS for vector similarity search. The SDK is exposed to agents via CodeIndexToolsMixin, which is already composed into the flagship GaiaAgent.

Installation

uv pip install -e '.[rag]'

FAISS and numpy live under the [rag] extras. If they’re missing the SDK raises:

code_index dependencies missing. Install with: pip install -e '.[rag]'

CodeIndexConfig

from gaia.code_index.sdk import CodeIndexConfig

config = CodeIndexConfig(
    repo_path=".",
    max_files=5000,
    max_file_size_mb=1.0,
    chunk_overlap=50,
    embedding_model="user.embeddinggemma-300m-GGUF",
    cache_dir="~/.gaia/code_index",
    embedding_base_url=None,
)

Field	Type	Default	Description
`repo_path`	`str`	—	Repository root (required)
`max_files`	`int`	`5000`	Max files to scan
`max_file_size_mb`	`float`	`1.0`	Skip files larger than this
`chunk_overlap`	`int`	`50`	Token overlap between chunks
`embedding_model`	`str`	`"user.embeddinggemma-300m-GGUF"`	Lemonade embedding model
`cache_dir`	`str`	`"~/.gaia/code_index"`	Cache directory
`embedding_base_url`	`Optional[str]`	`None`	Custom Lemonade base URL

CodeIndexSDK

from gaia.code_index.sdk import CodeIndexSDK

sdk = CodeIndexSDK(config)

`index_repository() -> IndexResult`

Discover source files, parse them, embed via Lemonade, and persist a FAISS index. Re-runs are incremental — unchanged files (matched by SHA-256) reuse their existing embeddings.

result = sdk.index_repository()
print(f"Files: {result.files_indexed}, "
      f"Chunks: {result.chunks_created}, "
      f"Took: {result.duration_seconds:.1f}s")

`search(query, scope="all", top_k=10) -> List[SearchResult]`

results = sdk.search("error handling in agent tools", top_k=5)
for r in results:
    print(f"[{r.score:.3f}] {r.chunk.file_path}:{r.chunk.start_line} "
          f"— {r.chunk.symbol_name or '?'}")

Parameter	Type	Default	Description
`query`	`str`	—	Natural language or code query
`scope`	`str`	`"all"`	`"all"` or `"code"` (reserved for future filters)
`top_k`	`int`	`10`	Number of results

L2 distance is converted to a similarity score in [0, 1] via 1 / (1 + dist).

`get_status() -> Dict`

status = sdk.get_status()
# {
#   "indexed": True,
#   "repo_path": "/abs/path",
#   "embedding_model": "user.embeddinggemma-300m-GGUF",
#   "total_chunks": 1234,
#   "code_chunks": 1234,
#   "files_tracked": 312,
#   "created_at": 1714000000.0,
#   "cache_path": "/Users/me/.gaia/code_index/<hash>"
# }

`clear_index()`

Removes the FAISS index and metadata for this repo from the cache.

sdk.clear_index()

Data types

CodeChunk

Field	Type	Description
`content`	`str`	Source text
`file_path`	`str`	Path relative to repo root
`language`	`str`	Detected language
`start_line`	`int`	First line (1-indexed)
`end_line`	`int`	Last line
`symbol_name`	`Optional[str]`	Function / class name
`symbol_type`	`Optional[str]`	`"function"`, `"class"`, …
`docstring`	`Optional[str]`	Extracted docstring
`imports`	`List[str]`	Imports parsed from the file

SearchResult

Field	Type	Description
`chunk`	`CodeChunk`	The matched chunk
`score`	`float`	Similarity score in `[0, 1]` (higher is better)
`result_type`	`str`	Currently always `"code"`

IndexResult

Field	Type	Description
`files_indexed`	`int`	Files newly parsed in this run
`chunks_created`	`int`	Total chunks in the resulting index (reused + new)
`duration_seconds`	`float`	Wall-clock time for the run

Parsers

gaia.code_index.parsers provides language detection and symbol-aware chunking.

from gaia.code_index.parsers import (
    detect_language,
    is_binary_file,
    parse_python_file,
    parse_generic_file,
    chunk_code_file,
)

# Detect language from filename
lang = detect_language("agent.py")  # "python"

# Parse a Python file with the AST
chunks = parse_python_file("agent.py", source_code)

# Parse JS/TS/Go/Rust/Java/C/C++ via regex
chunks = parse_generic_file("agent.ts", source_code, "typescript")

# High-level dispatcher
chunks = chunk_code_file("agent.go", source_code, max_size_mb=1.0)

Agent integration: CodeIndexToolsMixin

The mixin lives at gaia.agents.tools.code_index_tools.CodeIndexToolsMixin. It exposes four @tool-decorated methods through register_code_index_tools():

Tool	Description
`index_codebase(repo_path="")`	Build / refresh the FAISS index
`search_code_index(query, scope="all", top_k=10)`	Semantic search
`get_index_status()`	Report indexed state, chunk counts, embedding model, cache path
`clear_code_index()`	Remove the cached index

GaiaAgent already composes this mixin, so the flagship agent’s coding skill gets the tools for free — see the Code Index guide.

Composing onto a custom Python agent

from gaia.agents.base.agent import Agent
from gaia.agents.tools.code_index_tools import CodeIndexToolsMixin

class MyResearchAgent(CodeIndexToolsMixin, Agent):
    def __init__(self, repo_path: str = "."):
        super().__init__()
        self._init_code_index_state(repo_path=repo_path)

    def _register_tools(self):
        super()._register_tools()
        self.register_code_index_tools()

State is initialised lazily — calling _init_code_index_state is optional but recommended so the agent honours the repo path you intended. CodeIndexToolsMixin is also registered in KNOWN_TOOLS["code_index"] (src/gaia/agents/registry.py) for dynamic composition when scaffolding agents — see the custom-agent guide.

Privacy and safety

All embeddings are generated locally by Lemonade Server.
Sensitive filenames are skipped during discovery: .env, .env.*, SSH private keys, *.pem, *.key, *.pfx, *.p12, *.jks, *.keystore.
The SDK uses gaia.security.PathValidator to keep file reads scoped to the configured repo_path, and the index_codebase tool refuses any repo_path that isn’t a sub-path of the agent’s original root.

Roadmap

#869 — MCP server wrapper exposing the index to external code assistants.
#870 — Multi-repo indexing in a single namespace.
#871 — Verilog / VHDL parsers.

Getting Started

C++ Framework

Python Framework

Installation

CodeIndexConfig

CodeIndexSDK

`index_repository() -> IndexResult`

`search(query, scope="all", top_k=10) -> List[SearchResult]`

`get_status() -> Dict`

`clear_index()`

Data types

CodeChunk

SearchResult

IndexResult

Parsers

Agent integration: CodeIndexToolsMixin

Composing onto a custom Python agent

Privacy and safety

Roadmap

​Installation

​CodeIndexConfig

​CodeIndexSDK

​index_repository() -> IndexResult

​search(query, scope="all", top_k=10) -> List[SearchResult]

​get_status() -> Dict

​clear_index()

​Data types

​CodeChunk

​SearchResult

​IndexResult

​Parsers

​Agent integration: CodeIndexToolsMixin

​Composing onto a custom Python agent

​Privacy and safety

​Roadmap

Installation

CodeIndexConfig

CodeIndexSDK

`index_repository() -> IndexResult`

`search(query, scope="all", top_k=10) -> List[SearchResult]`

`get_status() -> Dict`

`clear_index()`

Data types

CodeChunk

SearchResult

IndexResult

Parsers

Agent integration: CodeIndexToolsMixin

Composing onto a custom Python agent

Privacy and safety

Roadmap