models

models

Simple data classes — no numpy imports to avoid JIT crashes.

AgentResult module-attribute

AgentResult = namedtuple(
    "AgentResult",
    [
        "answer",
        "sources",
        "chunks",
        "tools_called",
        "search_used",
    ],
)

Result from a synchronous agent process() call.

Fields

answer: The agent's final text response. sources: List of source citation tags (e.g., [source: file.md]). chunks: List of ChunkResult objects from RAG retrieval. tools_called: List of dicts with 'name' and 'args' for each tool invoked. search_used: True if Tavily web search was used during this turn.

Chunk

A single text chunk from the knowledge base.

Represents one segment of a source document, optionally scoped to a specific section heading. Chunks are the unit of retrieval and citation in RAG queries.

Attributes:

Name	Type	Description
text		The chunk's text content.
source		Source file path this chunk originated from.
section		Section heading under which this chunk falls (may be empty).

__init__

__init__(text: str, source: str, section: str = '')

Initialize a Chunk.

Parameters:

Name	Type	Description	Default
text	str	The chunk's text content.	required
source	str	Source file path this chunk originated from.	required
section	str	Section heading under which this chunk falls (may be empty).	''

ChunkResult

A ranked retrieval result from a knowledge base query.

Returned by KnowledgeBase.query() and rag_retrieve(). Carries the matched chunk's content, its source file, and a cosine similarity score against the query vector.

Attributes:

Name	Type	Description
text		The matched chunk's text content.
source		Source file path of the chunk.
score		Cosine similarity score (0.0 to 1.0 range).

__init__

__init__(text: str, source: str, score: float)

Initialize a ChunkResult.

Parameters:

Name	Type	Description	Default
text	str	The matched chunk's text content.	required
source	str	Source file path of the chunk.	required
score	float	Cosine similarity score (0.0 to 1.0 range).	required

ChatRequest dataclass

Request body for the /chat and /chat/stream endpoints.

Attributes:

Name	Type	Description
message	str	The user's chat message or question.
session_id	Optional[str]	Optional UUID of an existing session to continue.
axes	Optional[dict]	Optional dict overriding the global axis values for this request.

ChatResponse dataclass

Response body for the non-streaming /chat endpoint.

Attributes:

Name	Type	Description
reply	str	The agent's complete text response.
session_id	Optional[str]	UUID of the session this turn belongs to.
sources	list	List of source filenames cited in the response.

LoginRequest dataclass

Request body for the /auth/login endpoint.

Attributes:

Name	Type	Description
email	str	User's email address.
password	str	User's plaintext password.

LoginResponse dataclass

Response body for successful /auth/login requests.

Attributes:

Name	Type	Description
access_token	str	JWT access token (15-minute expiry).
token_type	str	Token type, always 'bearer'.