MCP Server for Langfuse Prompts

# Langfuse Prompt Management Migration

## Goal

Your goal is to identify hardcoded prompts in this codebase and migrate them to Langfuse Prompt Management, then update the code to fetch prompts from Langfuse at runtime.

## Rules

Before you begin, understand these fundamental rules:

1.  **Preserve Business Logic**: Do not change application behavior. The only changes should be moving prompts to Langfuse and updating how prompts are retrieved.
2.  **Follow the Workflow**: Execute the steps below in sequence.
3.  **Use Available Tools**: Use the Langfuse MCP server tools (`listPrompts`, `getPrompt`, `createTextPrompt`, `createChatPrompt`, `updatePromptLabels`) to interact with Langfuse Prompt Management.

## Migration Workflow

### Step 1: Discover Hardcoded Prompts

Analyze the codebase to find all hardcoded prompts. Look for:

- System prompts or instructions passed to LLM calls
- Template strings used for prompt construction
- Multi-line strings containing AI instructions
- Variables named `prompt`, `system_prompt`, `instructions`, `template`, etc.
- Files in directories like `prompts/`, `templates/`, or similar

Present a list of all discovered prompts with:
- File path and line numbers
- The prompt content (truncated if very long)
- Suggested prompt name for Langfuse (e.g., `customer-support-agent`, `code-review-assistant`)
- Prompt type: text or chat (OpenAI message format)

Ask me to confirm which prompts to migrate before proceeding.

### Step 2: Check Existing Prompts in Langfuse

Use the `listPrompts` tool to check what prompts already exist in the Langfuse project. Compare with the prompts identified in Step 1 to avoid duplicates.

### Step 3: Create Prompts in Langfuse

For each confirmed prompt, create it in Langfuse:

**For simple text prompts** (single string templates):
- Use `createTextPrompt` tool
- Extract template variables and use `{{variable}}` syntax
- Add appropriate tags (e.g., `migrated`, `v1`)
- Set the `production` label for immediate use

**For chat prompts** (system/user/assistant message format):
- Use `createChatPrompt` tool
- Structure as OpenAI-style messages with roles
- Include all message types (system, user examples, etc.)
- Add appropriate tags and labels

Include a commit message describing the migration (e.g., "Migrated from codebase: path/to/file.py")

### Step 4: Propose Code Changes

Present a plan to update the codebase to fetch prompts from Langfuse at runtime. The plan should include:

1.  **SDK Installation**: The Langfuse SDK package to install
2.  **Environment Variables**: Required configuration (LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY, LANGFUSE_HOST if self-hosted)
3.  **Code Changes**: For each migrated prompt, show the before/after code

Wait for my approval before implementing changes.

### Step 5: Implement Code Changes

After approval, make the following changes:

1.  Install the Langfuse SDK using the package manager
2.  Add prompt fetching code that:
    - Initializes the Langfuse client
    - Fetches prompts by name (optionally with label for staging/production)
    - Compiles prompts with any required variables
    - Handles caching appropriately for your use case

**Python example pattern:**
```python
from langfuse import Langfuse

langfuse = Langfuse()

# Fetch and compile prompt
prompt = langfuse.get_prompt("prompt-name")
compiled = prompt.compile(variable="value")
```

**JavaScript/TypeScript example pattern:**
```typescript
import { Langfuse } from "langfuse";

const langfuse = new Langfuse();

// Fetch and compile prompt  
const prompt = await langfuse.getPrompt("prompt-name");
const compiled = prompt.compile({ variable: "value" });
```

3.  Remove or comment out the old hardcoded prompts
4.  Add documentation comments explaining the Langfuse prompt source

### Step 6: Verify Migration

After implementing changes:

1.  Confirm all prompts are created in Langfuse using `listPrompts`
2.  Verify prompt content is correct using `getPrompt`
3.  Ask me to test the application and confirm prompts are being fetched correctly

### Step 7: Optional Enhancements

Suggest additional improvements:

- **Labels**: Set up staging/production labels for A/B testing
- **Versioning**: Explain how to iterate on prompts via Langfuse UI
- **Caching**: Configure appropriate cache TTL for your use case
- **Fallbacks**: Add fallback behavior if Langfuse is unreachable