perplexity_search_web

Instructions

Implementation Reference

• src/perplexity_mcp/server.py:100-109 (handler)

  The MCP tool call handler that specifically handles 'perplexity_search_web' by parsing arguments and delegating to the API call helper.

  @server.call_tool()
  async def call_tool(
      name: str, arguments: dict
  ) -> list[types.TextContent | types.ImageContent | types.EmbeddedResource]:
      if name == "perplexity_search_web":
          query = arguments["query"]
          recency = arguments.get("recency", "month")
          result = await call_perplexity(query, recency)
          return [types.TextContent(type="text", text=str(result))]
      raise ValueError(f"Tool not found: {name}")

• src/perplexity_mcp/server.py:112-157 (helper)

  Core helper function that performs the HTTP POST to Perplexity AI API, applies recency filter via search_recency_filter, and formats response with citations.

  async def call_perplexity(query: str, recency: str) -> str:
  
      url = "https://api.perplexity.ai/chat/completions"
  
      # Get the model from environment variable or use "sonar" as default
      model = os.getenv("PERPLEXITY_MODEL", "sonar")
  
      payload = {
          "model": model,
          "messages": [
              {"role": "system", "content": "Be precise and concise."},
              {"role": "user", "content": query},
          ],
          "max_tokens": "512",
          "temperature": 0.2,
          "top_p": 0.9,
          "return_images": False,
          "return_related_questions": False,
          "search_recency_filter": recency,
          "top_k": 0,
          "stream": False,
          "presence_penalty": 0,
          "frequency_penalty": 1,
          "return_citations": True,
          "search_context_size": "low",
      }
  
      headers = {
          "Authorization": f"Bearer {os.getenv('PERPLEXITY_API_KEY')}",
          "Content-Type": "application/json",
      }
  
      async with aiohttp.ClientSession() as session:
          async with session.post(url, json=payload, headers=headers) as response:
              response.raise_for_status()
              data = await response.json()
              content = data["choices"][0]["message"]["content"]
              
              # Format response with citations if available
              if "citations" in data:
                  citations = data["citations"]
                  formatted_citations = "\n\nCitations:\n" + "\n".join(f"[{i+1}] {url}" for i, url in enumerate(citations))
                  return content + formatted_citations
              
              return content

• src/perplexity_mcp/server.py:79-98 (registration)

  Registers the 'perplexity_search_web' tool with MCP server via list_tools decorator, including description and input schema.

  async def list_tools() -> list[types.Tool]:
      return [
          types.Tool(
              name="perplexity_search_web",
              description="Search the web using Perplexity AI with recency filtering",
              inputSchema={
                  "type": "object",
                  "properties": {
                      "query": {"type": "string"},
                      "recency": {
                          "type": "string",
                          "enum": ["day", "week", "month", "year"],
                          "default": "month",
                      },
                  },
                  "required": ["query"],
              },
          )
      ]

• src/perplexity_mcp/server.py:84-95 (schema)

  JSON schema defining tool inputs: query (required string), recency (optional enum: day, week, month, year with default month).

  inputSchema={
      "type": "object",
      "properties": {
          "query": {"type": "string"},
          "recency": {
              "type": "string",
              "enum": ["day", "week", "month", "year"],
              "default": "month",
          },
      },
      "required": ["query"],
  },