Structs§
- A2aConfig
- Host-owned A2A peer trust. Graphs and agent artifacts can select only an exact endpoint listed here; credentials never enter graph or artifact data.
- A2aPeer
Config - One exact remote UAR agent binding. Reusing a token for several agents is explicit: each endpoint still needs its own entry and artifact dependency.
- AcpConfig
- interface for agent introspection, session management, and streaming run execution — used by IDEs, debuggers, and the BeeAI platform.
- AppConfig
- Chunking
Config - Chunking strategy configuration.
- Cli
- Embedding
Backend Config - Embedding backend configuration for RAG and memory.
- Failover
Config - Runtime model failover configuration.
- Fallback
Model - A single entry in the ordered fallback model list.
- File
Processing Config - Configuration for file processing and uploads.
- Guardrails
Config - Input guardrail configuration: heuristic prompt-injection and PII/secret screening of chat input before the LLM call.
- Harness
Config - Knowledge
Base Config - Configuration for a single knowledge base.
- Knowledge
Bases Config - Top-level configuration for knowledge bases.
- Kreuzberg
Chunking Config - Kreuzberg text-chunking configuration for RAG pipelines. Field names use Rust conventions (max_characters/overlap) — differs from the Python SDK (max_chars/max_overlap). See kreuzberg SKILL.md pitfalls.
- Kreuzberg
Config - Kreuzberg local file processing configuration. Kreuzberg is a high-performance document intelligence framework with a Rust core.
- LlmBudget
Config - Budget enforcement configuration.
- LlmCache
Config - Cache configuration for the liter-llm Tower middleware.
- LlmConfig
- Unified LLM configuration for the liter-llm client.
- LlmRate
Limit Config - Rate limiting configuration.
- Memory
Config - Configuration for the in-process agent memory system, backed by surreal-memory + SurrealDB/SurrealKV.
- Mistral
Config - Mistral OCR configuration.
- Models
Config - Model files configuration (tokenizer, embeddings, etc.).
- Native
Tools Config - Configuration for native (in-process) tool implementations.
- Persistence
Config - Resilience
Config - Runs
Config - Sandbox
Runtime Config - Configuration for the sandboxed code-execution subsystem.
- Security
Config - Server
Config - Sessions
Config - Skill
Evolution Config - Configuration for the post-run skill evolution system.
- Skill
Reattachment Budget - Model-token ceilings for re-attaching active bodies after compaction.
- Sycophancy
Config - Configuration for the ACP (Agent Communication Protocol) server endpoint.
- Unstructured
Config - Unstructured.io configuration (hosted or self-hosted).
- Vision
Config - Vision/Image processing configuration.
Enums§
- Command
- Top-level binary subcommands. Absent ⇒ run the server (default).
- Eval
Action evalsubcommands.- Failover
Strategy - Selection strategy for picking the next provider when the primary fails.
- Harness
Mode - Typed assembly is the default after the recorded parity-and-smoke gate.
- LogFormat
- Reasoning
Effort - Host-selected reasoning level for one run.
- Skill
Activation Mode - Skill matching suggests catalog entries or activates legacy overlays.
Functions§
- build_
client_ config - Build a
liter_llm::ClientConfigfrom the mergedLlmConfig. - load_
llm_ config - Load just the merged
LlmConfigfrom all layered configuration sources.