Modules
Builtin generic action implementations.
A generic action impl that returns structured outputs from an LLM matching the action return.
Internal ReqLLM-native flow state for prompt actions.
A change that vectorizes the current values of attributes for a given record. Used by the manual vectorization strategy, so you can decide later when to run the vectorization action.
Vectorizes attributes inline immediately after they are changed
Run an ash_oban trigger when embeddings need to be regenerated.
Normalizes tool call and tool result data for generated chat UIs.
A check that is true when the actor is %AshAi{}
Spark DSL schemas and configuration for AshAi.
A behaviour that defines the dimensions of the vector, and how to generate the embedding
ReqLLM-backed embedding model for AshAi vectorization.
A section that defines how complex vectorized columns are defined
Introspection functions for the AshAi extension.
Model Context Protocol (MCP) implementation for Ash Framework.
Place in your endpoint's code_reloading section to expose Ash dev MCP"
MCP Router implementing the RPC functionality over HTTP.
Implementation of the Model Context Protocol (MCP) RPC functionality.
An MCP resource to expose via the Model Context Protocol (MCP).
A UI resource for MCP Apps (Model Context Protocol Apps extension).
An action exposed to LLM agents
A struct representing an argument defined in the Tool DSL.
Builds ReqLLM.Tool structs and callbacks from AshAi.Tool DSL entities.
Formats Ash errors as JSON:API error objects for tool responses.
Executes Ash actions from tool call arguments.
Execution context for tool calls.
Generates JSON schemas for tool parameters.
Event data passed to the on_tool_end callback used by AshAi.ToolLoop.
Manages a ReqLLM conversation loop with tool calls.
Event emitted at the start of each iteration in the tool loop.
Result returned from a completed tool loop.
Event data passed to the on_tool_start callback used by AshAi.ToolLoop.
High-level API for discovering and building tools from Ash domains.
A validation that passes if the actor is %AshAi{}
Verifies that all MCP resource actions return strings.
Mix Tasks
Generates the resources and views for a conversational UI backed by ash_postgres and ash_oban
Sets up an MCP server for your application
Installs AshAi. Call with mix igniter.install ash_ai. Requires igniter to run.