Agent Langchain¶

Agent LangChain — mvp.agent_langchain

Cluster: Agents & LLM | Type: component | MCP Tools: 30

Overview¶

LangChain multi-provider chat agent that routes requests to OpenAI-compatible, OpenRouter, Anthropic, or Ollama backends through a unified interface. Supports multi-turn sessions via in-block state keyed by session_id, and integrates LifecycleMixin with resource guardrails for production use.

When to use:

Running LLM-backed conversations across OpenAI-compatible, OpenRouter, Anthropic, or local Ollama models without changing calling code
Maintaining persistent multi-turn conversation sessions within a single block instance
Embedding LangChain chains into the G6 pipeline alongside other AIBlock components

Example:

from mvp.agent_langchain import AgentLangChainBlock, LangChainInput, LangChainMessage

block = AgentLangChainBlock(name="langchain")
result = block.infer(LangChainInput(
    messages=[LangChainMessage(role="user", content="Summarise quantum entanglement.")],
    provider="ollama",
    model="gpt-oss:20b",
    session_id="sess-001",
))
# result.ok → True; result.value → LangChainOutput with response, provider, session_id

Works well with: agent_openai, agent_autogen, llm_router

Launch Readiness Notes¶

This component is suitable for Phase 0 pilot workflows where the goal is to let a non-technical user install the MCP package, call a chat/chain/RAG workflow, and receive a useful result quickly.

Known caveats before treating it as hardened paid-customer infrastructure:

The skill-level MCP delegate is thin; authoritative tool registration lives in langchain_mcp/server.py.
Provider capability discovery is explicit via provider_capabilities / langchain_provider_capabilities and reports booleans only; it must not expose API key values.
The MCP RAG path uses lightweight local TF-IDF retrieval and prompt assembly over caller/stored documents. It is useful for demos and small document sets, but it is not semantic/vector retrieval and is not robust enough for large corpora, citations, permissions, or compliance-sensitive retrieval.
The component relies on surrounding platform controls for authentication, rate limits, billing enforcement, tenant isolation, and production observability. Do not expose it as a standalone public endpoint without those layers.
Streaming responses are not implemented in this component path.

Public API¶

`AgentLangChainBlock(LifecycleMixin, AIBlock[LangChainInput, LangChainOutput, dict])`¶

LangChain multi-provider agent (anthropic / openrouter / ollama).

Field	Type	Default
`name`	`str`	'agent_langchain'
`state`	`dict`	field(default_factory=dict)
`resource_bounds`	`ResourceBounds \\| None`	None
`usage`	`ResourceUsage`	field(default_factory=ResourceUsage)

Methods:

`infer(data: LangChainInput) -> Result[LangChainOutput]`¶

`LangChainMessage(BaseModel)`¶

Field	Type	Default
`role`	`str`	required
`content`	`str`	required

`LangChainInput(BaseModel)`¶

Field	Type	Default
`messages`	`list[LangChainMessage]`	required
`system_prompt`	`str`	''
`model`	`str`	'gpt-4o-mini'
`max_tokens`	`int`	1024
`temperature`	`float`	0.7
`api_key`	`str`	''
`base_url`	`str`	''
`provider`	`str`	'openai'
`session_id`	`str`	''
`allow_paid_api`	`bool`	False

`LangChainOutput(BaseModel)`¶

Field	Type	Default
`response`	`str`	required
`model`	`str`	''
`metadata`	`dict[str, Any]`	Field(default_factory=dict)
`n_messages`	`int`	0
`provider`	`str`	''
`session_id`	`str`	''
`chain_steps`	`list[dict[str, Any]]`	Field(default_factory=list)
`degraded`	`bool`	False
`degradation_reason`	`str`	''
`completion_state`	`str`	'qualified-draft'
`warning_card`	`dict[str, Any]`	Field(default_factory=dict)
`evidence`	`dict[str, Any]`	Field(default_factory=dict)
`request_id`	`str`	''
`task_id`	`str`	''
`run_id`	`str`	''

`AdaptLangChainMCPBlock(AIBlock[MCPLangChainInput, MCPLangChainOutput, None])`¶

SQLite-backed LangChain MCP block with 27 ops.

Field	Type	Default
`name`	`str`	'AdaptLangChainMCPBlock'
`db_path`	`str`	':memory:'
`agentic_planner`	`object`	None

Methods:

`infer(data: MCPLangChainInput) -> Result[MCPLangChainOutput]`¶

`MCPLangChainInput(BaseModel)`¶

Field	Type	Default
`op`	`Literal['chat', 'chain', 'tools', 'rag', 'status', 'ops', 'help', 'save_session', 'load_session', 'list_sessions', 'delete_session', 'store_chain', 'retrieve_chain', 'list_chains', 'delete_chain', 'add_documents', 'search_documents', 'list_documents', 'delete_documents', 'register_tool', 'list_tools', 'remove_tool', 'record_interaction', 'query_interactions', 'summarize_interactions', 'search', 'info', 'provider_capabilities', 'recommend_provider', 'list_patterns']`	required
`messages`	`list[dict]`	Field(default_factory=list)
`system_prompt`	`str`	''
`prompt_template`	`str`	''
`variables`	`dict[str, Any]`	Field(default_factory=dict)
`tools`	`list[dict]`	Field(default_factory=list)
`query`	`str`	''
`documents`	`list[str]`	Field(default_factory=list)
`top_k`	`int`	3
`model`	`str`	'gpt-4o-mini'
`max_tokens`	`int`	1024
`temperature`	`float`	0.7
`api_key`	`str`	''
`base_url`	`str`	''
`provider`	`str`	''
`allow_paid_api`	`bool`	False
`session_id`	`str`	''
`session_name`	`str`	''
`metadata`	`dict[str, Any]`	Field(default_factory=dict)
`chain_id`	`str`	''
`chain_name`	`str`	''
`parser`	`str`	'str'
`config`	`dict[str, Any]`	Field(default_factory=dict)
`doc_id`	`str`	''
`content`	`str`	''
`source`	`str`	''
`tool_name`	`str`	''
`description`	`str`	''
`tool_schema`	`dict[str, Any]`	Field(default_factory=dict, alias='schema')
`input_data`	`dict[str, Any]`	Field(default_factory=dict)
`response`	`str`	''
`limit`	`int`	20

`MCPLangChainOutput(BaseModel)`¶

Field	Type	Default
`op`	`str`	''
`error`	`str`	''
`key`	`str`	''
`value`	`Any`	None
`found`	`bool`	False
`count`	`int`	0
`records`	`list[dict]`	Field(default_factory=list)
`retrieved`	`list[Any]`	Field(default_factory=list)
`summary`	`str`	''
`message`	`str`	''
`metadata`	`dict[str, Any]`	Field(default_factory=dict)
`degraded`	`bool`	False
`degradation_reason`	`str`	''
`completion_state`	`str`	'qualified-draft'
`warning_card`	`dict[str, Any]`	Field(default_factory=dict)
`evidence`	`dict[str, Any]`	Field(default_factory=dict)
`request_id`	`str`	''
`task_id`	`str`	''
`run_id`	`str`	''

`LangChainStore`¶

Constructor:

Parameter	Type	Default
`db_path`	`str`	':memory:'

Methods:

`save_session(session_id: str, session_name: str, messages: list, model: str, metadata: dict) -> dict`¶

`load_session(session_id: str) -> dict | None`¶

`list_sessions() -> list[dict]`¶

`delete_session(session_id: str) -> int`¶

`store_chain(chain_id: str, chain_name: str, prompt_template: str, model: str, parser: str, config: dict) -> dict`¶

`retrieve_chain(chain_id: str) -> dict | None`¶

`list_chains() -> list[dict]`¶

`delete_chain(chain_id: str) -> int`¶

`add_documents(doc_id: str, content: str, source: str, metadata: dict) -> dict`¶

`list_documents() -> list[dict]`¶

`delete_documents(doc_id: str) -> int`¶

`search_documents(query: str, top_k: int = 3) -> list[dict]`¶

`register_tool(tool_name: str, description: str, schema: dict) -> dict`¶

`list_tools() -> list[dict]`¶

`remove_tool(tool_name: str) -> int`¶

`record_interaction(session_id: str, input_data: dict, response: str, model: str, metadata: dict) -> dict`¶

`query_interactions(session_id: str = '', limit: int = 20) -> list[dict]`¶

`summarize_interactions(session_id: str = '') -> dict`¶

`search(query: str, top_k: int = 5) -> list[dict]`¶

`count_all() -> dict`¶

`info() -> dict`¶

MCP Tools¶

Operation	Source
`chat`	langchain_mcp
`chain`	langchain_mcp
`tools`	langchain_mcp
`rag`	langchain_mcp
`status`	langchain_mcp
`ops`	langchain_mcp
`help`	langchain_mcp
`save_session`	langchain_mcp
`load_session`	langchain_mcp
`list_sessions`	langchain_mcp
`delete_session`	langchain_mcp
`store_chain`	langchain_mcp
`retrieve_chain`	langchain_mcp
`list_chains`	langchain_mcp
`delete_chain`	langchain_mcp
`add_documents`	langchain_mcp
`search_documents`	langchain_mcp
`list_documents`	langchain_mcp
`delete_documents`	langchain_mcp
`register_tool`	langchain_mcp
`list_tools`	langchain_mcp
`remove_tool`	langchain_mcp
`record_interaction`	langchain_mcp
`query_interactions`	langchain_mcp
`summarize_interactions`	langchain_mcp
`search`	langchain_mcp
`info`	langchain_mcp
`provider_capabilities`	langchain_mcp
`recommend_provider`	langchain_mcp
`list_patterns`	langchain_mcp

Agent Langchain¶

Overview¶

Launch Readiness Notes¶

Public API¶

AgentLangChainBlock(LifecycleMixin, AIBlock[LangChainInput, LangChainOutput, dict])¶

infer(data: LangChainInput) -> Result[LangChainOutput]¶

LangChainMessage(BaseModel)¶

LangChainInput(BaseModel)¶

LangChainOutput(BaseModel)¶

AdaptLangChainMCPBlock(AIBlock[MCPLangChainInput, MCPLangChainOutput, None])¶

infer(data: MCPLangChainInput) -> Result[MCPLangChainOutput]¶

MCPLangChainInput(BaseModel)¶

MCPLangChainOutput(BaseModel)¶

LangChainStore¶

save_session(session_id: str, session_name: str, messages: list, model: str, metadata: dict) -> dict¶

load_session(session_id: str) -> dict | None¶

list_sessions() -> list[dict]¶

delete_session(session_id: str) -> int¶

store_chain(chain_id: str, chain_name: str, prompt_template: str, model: str, parser: str, config: dict) -> dict¶

retrieve_chain(chain_id: str) -> dict | None¶

list_chains() -> list[dict]¶

delete_chain(chain_id: str) -> int¶

add_documents(doc_id: str, content: str, source: str, metadata: dict) -> dict¶

list_documents() -> list[dict]¶

delete_documents(doc_id: str) -> int¶

search_documents(query: str, top_k: int = 3) -> list[dict]¶

register_tool(tool_name: str, description: str, schema: dict) -> dict¶

list_tools() -> list[dict]¶

remove_tool(tool_name: str) -> int¶

record_interaction(session_id: str, input_data: dict, response: str, model: str, metadata: dict) -> dict¶

query_interactions(session_id: str = '', limit: int = 20) -> list[dict]¶

summarize_interactions(session_id: str = '') -> dict¶

search(query: str, top_k: int = 5) -> list[dict]¶

count_all() -> dict¶

info() -> dict¶

MCP Tools¶

`AgentLangChainBlock(LifecycleMixin, AIBlock[LangChainInput, LangChainOutput, dict])`¶

`infer(data: LangChainInput) -> Result[LangChainOutput]`¶

`LangChainMessage(BaseModel)`¶

`LangChainInput(BaseModel)`¶

`LangChainOutput(BaseModel)`¶

`AdaptLangChainMCPBlock(AIBlock[MCPLangChainInput, MCPLangChainOutput, None])`¶

`infer(data: MCPLangChainInput) -> Result[MCPLangChainOutput]`¶

`MCPLangChainInput(BaseModel)`¶

`MCPLangChainOutput(BaseModel)`¶

`LangChainStore`¶

`save_session(session_id: str, session_name: str, messages: list, model: str, metadata: dict) -> dict`¶

`load_session(session_id: str) -> dict | None`¶

`list_sessions() -> list[dict]`¶

`delete_session(session_id: str) -> int`¶

`store_chain(chain_id: str, chain_name: str, prompt_template: str, model: str, parser: str, config: dict) -> dict`¶

`retrieve_chain(chain_id: str) -> dict | None`¶

`list_chains() -> list[dict]`¶

`delete_chain(chain_id: str) -> int`¶

`add_documents(doc_id: str, content: str, source: str, metadata: dict) -> dict`¶

`list_documents() -> list[dict]`¶

`delete_documents(doc_id: str) -> int`¶

`search_documents(query: str, top_k: int = 3) -> list[dict]`¶

`register_tool(tool_name: str, description: str, schema: dict) -> dict`¶

`list_tools() -> list[dict]`¶

`remove_tool(tool_name: str) -> int`¶

`record_interaction(session_id: str, input_data: dict, response: str, model: str, metadata: dict) -> dict`¶

`query_interactions(session_id: str = '', limit: int = 20) -> list[dict]`¶

`summarize_interactions(session_id: str = '') -> dict`¶

`search(query: str, top_k: int = 5) -> list[dict]`¶

`count_all() -> dict`¶

`info() -> dict`¶