ask_perplexity
Gather source-backed information from the internet for research, fact-checking, or contextual data, with citations for transparent references.
Instructions
Perplexity equips agents with a specialized tool for efficiently gathering source-backed information from the internet, ideal for scenarios requiring research, fact-checking, or contextual data to inform decisions and responses. Each response includes citations, which provide transparent references to the sources used for the generated answer, and choices, which contain the model's suggested responses, enabling users to access reliable information and diverse perspectives. This function may encounter timeout errors due to long processing times, but retrying the operation can lead to successful completion. [Response structure]
id: An ID generated uniquely for each response.
model: The model used to generate the response.
object: The object type, which always equals
chat.completion.created: The Unix timestamp (in seconds) of when the completion was created.
citations[]: Citations for the generated answer.
choices[]: The list of completion choices the model generated for the input prompt.
usage: Usage statistics for the completion request.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| model | Yes | The name of the model that will complete your prompt. | |
| messages | Yes | A list of messages comprising the conversation so far. |
Implementation Reference
- The `handle_call_tool` function serves as the handler for the "ask_perplexity" tool. It validates the tool name, makes an asynchronous POST request to the Perplexity API's chat/completions endpoint using the provided arguments, handles errors, and returns the API response as text content.
@server.call_tool() async def handle_call_tool( name: str, arguments: dict ) -> list[types.TextContent | types.ImageContent | types.EmbeddedResource]: if name != "ask_perplexity": raise ValueError(f"Unknown tool: {name}") try: async with httpx.AsyncClient() as client: response = await client.post( f"{PERPLEXITY_API_BASE_URL}/chat/completions", headers={ "Authorization": f"Bearer {PERPLEXITY_API_KEY}", "Content-Type": "application/json", }, json=arguments, timeout=None, ) response.raise_for_status() except httpx.HTTPError as e: raise RuntimeError(f"API error: {str(e)}") return [ types.TextContent( type="text", text=response.text, ) ] - The inputSchema defines the expected parameters for the "ask_perplexity" tool: a 'model' string (enum with available Perplexity models) and a 'messages' array of objects with 'content' and 'role'.
inputSchema={ "type": "object", "properties": { "model": { "type": "string", "description": "The name of the model that will complete your prompt.", "enum": [ "llama-3.1-sonar-small-128k-online", # Commenting out larger models,which have higher risks of timing out, # until Claude Desktop can handle long-running tasks effectively. # "llama-3.1-sonar-large-128k-online", # "llama-3.1-sonar-huge-128k-online", ], }, "messages": { "type": "array", "description": "A list of messages comprising the conversation so far.", "items": { "type": "object", "properties": { "content": { "type": "string", "description": "The contents of the message in this turn of conversation.", }, "role": { "type": "string", "description": "The role of the speaker in this turn of conversation. After the (optional) system message, user and assistant roles should alternate with user then assistant, ending in user.", "enum": ["system", "user", "assistant"], }, }, "required": ["content", "role"], }, }, }, "required": ["model", "messages"], }, - src/mcp_server_perplexity/server.py:20-83 (registration)The tool is registered in the `handle_list_tools` function via the `types.Tool` object with name="ask_perplexity", a detailed description, and the inputSchema.
types.Tool( name="ask_perplexity", description=dedent( """ Perplexity equips agents with a specialized tool for efficiently gathering source-backed information from the internet, ideal for scenarios requiring research, fact-checking, or contextual data to inform decisions and responses. Each response includes citations, which provide transparent references to the sources used for the generated answer, and choices, which contain the model's suggested responses, enabling users to access reliable information and diverse perspectives. This function may encounter timeout errors due to long processing times, but retrying the operation can lead to successful completion. [Response structure] - id: An ID generated uniquely for each response. - model: The model used to generate the response. - object: The object type, which always equals `chat.completion`. - created: The Unix timestamp (in seconds) of when the completion was created. - citations[]: Citations for the generated answer. - choices[]: The list of completion choices the model generated for the input prompt. - usage: Usage statistics for the completion request. """ ), inputSchema={ "type": "object", "properties": { "model": { "type": "string", "description": "The name of the model that will complete your prompt.", "enum": [ "llama-3.1-sonar-small-128k-online", # Commenting out larger models,which have higher risks of timing out, # until Claude Desktop can handle long-running tasks effectively. # "llama-3.1-sonar-large-128k-online", # "llama-3.1-sonar-huge-128k-online", ], }, "messages": { "type": "array", "description": "A list of messages comprising the conversation so far.", "items": { "type": "object", "properties": { "content": { "type": "string", "description": "The contents of the message in this turn of conversation.", }, "role": { "type": "string", "description": "The role of the speaker in this turn of conversation. After the (optional) system message, user and assistant roles should alternate with user then assistant, ending in user.", "enum": ["system", "user", "assistant"], }, }, "required": ["content", "role"], }, }, }, "required": ["model", "messages"], }, ) ]