CogneeWriter

Write ChatMessage objects into a Cognee knowledge graph memory.

Key Features

• Persists ChatMessage objects into a Cognee knowledge graph for long-term agent memory.
• Supports both the permanent knowledge graph (no session_id) and session-scoped caching.
• The writer's session_id overrides the store's session_id, so one store can back multiple writers targeting different memory tiers.
• Passes written messages through unchanged so they can continue to flow to downstream components.
• Works with CogneeRetriever to form a complete memory write/read pair in agent pipelines.

Configuration

1. Drag the CogneeWriter component onto the canvas from the Component Library.
2. Click on the component to open the configuration panel.
3. On the General tab:
   1. Configure the memory_store with a CogneeMemoryStore, setting the dataset_name and optionally session_id.
   2. Optionally set a session_id on the writer to override the store's session for this writer's writes.

Connections

CogneeWriter receives a list of ChatMessage objects and an optional user_id. It stores the messages in Cognee and passes them through unchanged via its messages_written output.

Use this component in agent pipelines to persist conversation history or other messages into Cognee's knowledge graph. Connect the source of messages (for example, an LLM reply) to its messages input.

Source Code

To check this component's source code, open memory_writer.py in the Haystack Core Integrations repository.

Usage Examples

Basic Configuration

  CogneeWriter:
    type: haystack_integrations.components.writers.cognee.memory_writer.CogneeWriter
    init_parameters:
      memory_store:
        type: haystack_integrations.memory_stores.cognee.memory_store.CogneeMemoryStore
        init_parameters:
          search_type: GRAPH_COMPLETION
          top_k: 5
          dataset_name: haystack_memory
          session_id:
          self_improvement: true
          timeout: 300
      session_id:

Using the Component in a Pipeline

This is an example of an agent pipeline that writes LLM replies into Cognee memory for future retrieval.

# haystack-pipeline
components:
  llm:
    type: haystack.components.generators.chat.openai.OpenAIChatGenerator
    init_parameters:
      api_key:
        type: env_var
        env_vars:
        - OPENAI_API_KEY
        strict: false
      model: gpt-4o-mini

  CogneeWriter:
    type: haystack_integrations.components.writers.cognee.memory_writer.CogneeWriter
    init_parameters:
      memory_store:
        type: haystack_integrations.memory_stores.cognee.memory_store.CogneeMemoryStore
        init_parameters:
          search_type: GRAPH_COMPLETION
          top_k: 5
          dataset_name: haystack_memory
          session_id:
          self_improvement: true
          timeout: 300
      session_id:

connections:
- sender: llm.replies
  receiver: CogneeWriter.messages

max_runs_per_component: 100

metadata: {}

inputs:
  messages:
  - llm.messages

outputs:
  messages_written: CogneeWriter.messages_written

Parameters

Inputs

Parameter	Type	Description
messages	List[ChatMessage]	The messages to persist into Cognee memory.
user_id	Optional[str]	Cognee user UUID to scope the write to a specific user.

Outputs

Parameter	Type	Description
messages_written	List[ChatMessage]	The same messages that were written, passed through unchanged.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
memory_store	CogneeMemoryStore		The backing CogneeMemoryStore to write into. Configure it with dataset_name and optionally session_id.
session_id	Optional[str]	None	Overrides the store's session_id for this writer's writes. Use to write to the session-cache tier instead of the permanent graph.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
messages	List[ChatMessage]		The messages to persist into Cognee memory.
user_id	Optional[str]	None	Cognee user UUID to scope the write to a specific user.

Related Information

• CogneeRetriever
• Pipeline Components