langchain: dynamic system prompt w/ middleware docs #592
Conversation
There was a problem hiding this comment.
Pull Request Overview
This PR adds documentation for dynamic system prompts using middleware in LangChain agents. The feature allows developers to modify system prompts dynamically at runtime based on context or agent state.
- Adds
DynamicSystemPromptMiddlewaredocumentation to the middleware page with comprehensive examples - Integrates dynamic prompt examples into the agents page to show practical usage
- Updates navigation and cross-references between middleware and agents documentation
Reviewed Changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| src/oss/langchain/middleware.mdx | Adds complete documentation section for DynamicSystemPromptMiddleware with examples for both context-based and conversation-length-based prompts |
| src/oss/langchain/agents.mdx | Adds a subsection on dynamic prompts with middleware to show integration with agent creation |
|
Preview ID generated: preview-srdyna-1758241143-49d6752 |
Co-authored-by: Christian Bromann <[email protected]>
Co-authored-by: Christian Bromann <[email protected]>
|
Preview ID generated: preview-srdyna-1758245605-9bcb701 |
Co-authored-by: Christian Bromann <[email protected]>
|
Preview ID generated: preview-srdyna-1758309488-c704e9e |
|
Preview ID generated: preview-srdyna-1758663482-f8412db |
|
Preview ID generated: preview-srdyna-1758663628-24b4adb |
christian-bromann
force-pushed
the
sr/dynamic-prompt
branch
from
5b4fb93 to
668b0f0
Compare
|
Preview ID generated: preview-srdyna-1758745243-7ad6f94 |
Co-authored-by: Copilot <[email protected]>
|
|
||
| ```python | ||
| from typing import TypedDict | ||
|
|
There was a problem hiding this comment.
The import order doesn't follow Python conventions. Standard library imports should come first, followed by third-party imports, then local imports. Move from typing import TypedDict to the top of the imports section.
Copilot uses AI. Check for mistakes.
| user_role: str | ||
|
|
||
| @modify_model_request | ||
| def dynamic_system_prompt(state: AgentState, request: ModelRequest, runtime: Runtime[Context]) -> ModelRequest: |
There was a problem hiding this comment.
The ModelRequest type is referenced but never imported or defined in the example. This will cause import errors when users try to run the code. Add the necessary import statement.
Copilot uses AI. Check for mistakes.
|
|
||
| const agent = createAgent({ | ||
| model: "openai:gpt-4o", | ||
| tools: [...], |
There was a problem hiding this comment.
The ellipsis [...] in the tools array is unclear and not valid TypeScript syntax. Replace with a comment like // your tools here or show actual tool examples for clarity.
| tools: [...], | |
| tools: [/* your tools here */], |
Copilot uses AI. Check for mistakes.
| user_role: str | ||
|
|
||
| @modify_model_request | ||
| def dynamic_system_prompt(state: AgentState, request: ModelRequest, runtime: Runtime[Context]) -> ModelRequest: |
There was a problem hiding this comment.
Similar to the middleware.mdx file, the ModelRequest type is used without being imported. Add the necessary import statement to make this example runnable.
Copilot uses AI. Check for mistakes.
|
|
||
| const agent = createAgent({ | ||
| model: "openai:gpt-4o", | ||
| tools: [/* ... */], |
There was a problem hiding this comment.
The comment /* ... */ in the tools array should be more descriptive. Consider using /* your tools here */ or provide an actual example tool to make the documentation more helpful.
| tools: [/* ... */], | |
| tools: [ | |
| { | |
| name: "calculator", | |
| description: "Performs basic arithmetic operations", | |
| func: async ({ a, b }) => a + b, | |
| }, | |
| ], |
Copilot uses AI. Check for mistakes.
|
Preview ID generated: preview-srdyna-1758745406-14f2e46 |
In Python we're now exposing decorators that can be used to generate simple middlewares w/ one hook.
TBD what we're doing in JS.