Knowledge Documents

BindAI V1 Knowledge uses project-local documents as reference information for Agents. The verified V1 flow is:

Knowledge Document
       |
       v
Project Knowledge
       |
       v
Agent Configuration
       |
       v
Agent Execution

Knowledge Documents

A Knowledge document contains information that an Agent may need during execution. Examples include:

Product information
Support information
Project documentation
Internal guides
Reference material

A project can contain Knowledge content that is configured for use by an Agent. For example:

my-project/
├── agents/
├── workflows/
├── tools/
└── knowledge/
    └── knowledge-test.md

Adding Knowledge

Knowledge is managed as project data. In Studio, Knowledge can be added to the project and then selected when configuring an Agent. The basic flow is:

Project
   |
   v
Knowledge
   |
   v
Agent Configuration
   |
   v
Select Knowledge
   |
   v
Workflow Execution

The Agent’s Knowledge configuration determines which project Knowledge is available during execution.

Concrete V1 Example

A verified V1 Knowledge example uses:

knowledge-test.md

with the content:

ORANGE-PENGUIN-742

The Agent is configured with:

Knowledge: knowledge-test

When the Agent executes, the Knowledge runtime can retrieve information from the configured Knowledge. The resulting flow is:

knowledge-test.md
       |
       v
knowledge-test
       |
       v
Agent
       |
       v
Knowledge search
       |
       v
ORANGE-PENGUIN-742

This verifies the document-to-Agent path implemented in V1.

Documents and Agents

A Knowledge document becomes useful to an Agent through the Agent’s Knowledge configuration. The runtime flow is:

Workflow
    |
    v
Agent
    |
    v
Configured Knowledge
    |
    v
Knowledge search
    |
    v
Relevant information
    |
    v
Model

The workflow does not need to copy the document contents into its input variables.

Documents and Memory

Knowledge and Memory have different responsibilities.

Knowledge	Memory
Project reference information	Persistent Agent conversation context
Project Knowledge	SQLite-backed project memory
Used during Agent execution	Loaded and saved around Agent execution
Available across workflow runs	Preserves conversation context

For example:

Knowledge
    |
    +-- Product information


Memory
    |
    +-- Previous conversation context

An Agent can use both during the same execution.

Documents and Workflow State

Knowledge documents are separate from workflow variables. Workflow state belongs to the current workflow execution:

Workflow State
    |
    +-- input
    +-- agent_output
    +-- tool_output
    +-- execution state

Knowledge belongs to the project:

Project
   |
   +-- Knowledge

A workflow run can therefore use project Knowledge without storing the document itself as workflow state.

Documents and Tools

Knowledge and Tools have different responsibilities. A Tool performs an operation:

Agent
  |
  v
Tool
  |
  v
Operation

Knowledge provides reference information:

Agent
  |
  v
Knowledge
  |
  v
Reference information

They can be combined in the same workflow:

Start
  |
  v
Agent
  |
  +-- Knowledge
  |
  v
Webhook Tool
  |
  v
End

The Agent can use Knowledge and then pass its result to a Tool through normal workflow variable mapping.

Documents and Conditions

A Knowledge-enabled Agent can participate in conditional workflows:

Start
  |
  v
Agent
  |
  v
Condition
  |
  +---- true ----> Tool
  |
  +---- false ---> Human Task

Knowledge provides information to the Agent. The Condition controls workflow routing.

Documents and Human Tasks

Knowledge can be used before a Human Task:

Start
  |
  v
Agent
  |
  v
Human Task
  |
  v
End

The Agent can use project Knowledge before the workflow pauses for human input. The Human Task controls the waiting and approval state.

Documents and Loops

A Knowledge-enabled Agent can participate in a Loop:

Start
  |
  v
Loop
  |
  v
Agent
  |
  v
End

The Loop controls repeated workflow execution. Knowledge provides reference information to the Agent.

Documents and Parallel Execution

Knowledge can be used by Agents in parallel branches:

             +---- Agent A ----+
             |                 |
Start -------+                 +---- Join
             |                 |
             +---- Agent B ----+

Each Agent uses the Knowledge configured for that Agent. Parallel and Join control execution and synchronization.

Updating Knowledge

When project reference information changes, the corresponding Knowledge content should be updated. For example:

Existing Knowledge
       |
       v
Updated project content
       |
       v
Future Agent executions

Knowledge should be kept current so Agents do not rely on outdated project information.

Removing Knowledge

When Knowledge is no longer required, it can be removed from the project through the available Knowledge management functionality. If an Agent references that Knowledge, its configuration should also be updated. The important distinction is:

Project Knowledge
       |
       +-- Available project reference information


Agent Configuration
       |
       +-- Which Knowledge the Agent uses

Document Content

Knowledge content should contain information relevant to the Agent. Useful content includes:

Product information
Support procedures
Project documentation
Frequently used reference information
Application-specific instructions

Keep Knowledge focused on information the Agent actually needs.

Document Quality

Good Knowledge content should be:

Clear
Relevant
Accurate
Maintained when the source information changes
Focused on the Agent’s purpose

For example, a support Agent should use support-related Knowledge rather than unrelated project information.

Document Security

Knowledge can contain private project information. Applications should consider:

Project access permissions
Filesystem permissions
Sensitive information
Backup security
Retention and deletion

Knowledge configuration itself is not an authorization system.

Testing Knowledge Documents

A useful V1 test verifies that an Agent can actually use the configured Knowledge. A basic test is:

Add Knowledge to the project.
Add known information to the Knowledge content.
Configure an Agent to use the Knowledge.
Execute the Agent through a workflow.
Ask about the known information.
Verify that the Agent can use the Knowledge.

The verified example is:

Knowledge:
    knowledge-test

Document:
    knowledge-test.md

Content:
    ORANGE-PENGUIN-742

The runtime should make that information available to the configured Agent.

Knowledge Documents and Memory

Knowledge and Memory can be used together:

Run 1
  |
  +-- Knowledge provides project information
  |
  +-- Memory stores conversation context


Run 2
  |
  +-- Knowledge remains available
  |
  +-- Memory loads previous conversation context

Knowledge provides project reference information. Memory provides persistent conversation context.

Current V1 Scope

Capability	V1
Project Knowledge	✅
Knowledge documents	✅
Knowledge management in Studio	✅
Agent Knowledge configuration	✅
Knowledge search during Agent execution	✅
Knowledge with Memory	✅
Knowledge with Tools	✅
Knowledge with Conditions	✅
Knowledge with Loops	✅
Knowledge with Parallel/Join	✅
Knowledge with Human Tasks	✅
Advanced document ingestion pipelines	❌
Custom embedding configuration	❌
BM25 configuration	❌
Hybrid retrieval configuration	❌
Reranker configuration	❌
Advanced indexing administration	❌
Advanced RAG pipeline administration	❌

Summary

BindAI V1 Knowledge documents provide project reference information that configured Agents can use during workflow execution. The practical V1 flow is:

Project Knowledge
       |
       v
Agent Configuration
       |
       v
Knowledge Search
       |
       v
Agent Context
       |
       v
Model
       |
       v
Workflow Result

Documents are separate from:

Memory, which provides persistent conversation context.
Workflow state, which belongs to the current workflow execution.
Tools, which perform operations.

The important V1 capability is concrete: project Knowledge can be configured for an Agent, searched during execution, and used by the Agent as reference information.

Getting Started

Core

Tools

Memory

Knowledge

Workflows

Projects

Templates

Connections

API Reference

Automation

Roadmap

05.2 documents

Knowledge Documents

Knowledge Documents

Adding Knowledge

Concrete V1 Example

Documents and Agents

Documents and Memory

Documents and Workflow State

Documents and Tools

Documents and Conditions

Documents and Human Tasks

Documents and Loops

Documents and Parallel Execution

Updating Knowledge

Removing Knowledge

Document Content

Document Quality

Document Security

Testing Knowledge Documents

Knowledge Documents and Memory

Current V1 Scope

Summary

​Knowledge Documents

​Knowledge Documents

​Adding Knowledge

​Concrete V1 Example

​Documents and Agents

​Documents and Memory

​Documents and Workflow State

​Documents and Tools

​Documents and Conditions

​Documents and Human Tasks

​Documents and Loops

​Documents and Parallel Execution

​Updating Knowledge

​Removing Knowledge

​Document Content

​Document Quality

​Document Security

​Testing Knowledge Documents

​Knowledge Documents and Memory

​Current V1 Scope

​Summary

Knowledge Documents

Knowledge Documents

Adding Knowledge

Concrete V1 Example

Documents and Agents

Documents and Memory

Documents and Workflow State

Documents and Tools

Documents and Conditions

Documents and Human Tasks

Documents and Loops

Documents and Parallel Execution

Updating Knowledge

Removing Knowledge

Document Content

Document Quality

Document Security

Testing Knowledge Documents

Knowledge Documents and Memory

Current V1 Scope

Summary