From 5ea048dc6ce02b223b02532b2d404a92fe637f5e Mon Sep 17 00:00:00 2001
From: Thomas Menard <menardorama@gmail.com>
Date: Sun, 3 May 2026 16:48:58 +0200
Subject: [PATCH] feat: support Firecrawl self-hosted/local instances via
 api_base config

FirecrawlFetchConfig now supports custom api_base URLs for:
- Self-hosted Firecrawl instances
- Local development
- On-premise deployments

Priority (highest to lowest):
1. api_base passed to afetch_url(optional_params)
2. api_base stored during validate_environment() from router config
3. FIRECRAWL_API_BASE environment variable
4. Default: https://api.firecrawl.dev/v1

Usage in proxy_config.yaml:
  fetch_tools:
    - fetch_tool_name: "my-local-firecrawl"
      litellm_params:
        fetch_provider: "firecrawl"
        api_base: "http://localhost:3002/api/v1"
        api_key: "os.environ/FIRECRAWL_API_KEY"

Also updated handler.py to pass api_base via optional_params.
---
 .../llms/firecrawl/fetch/transformation.py    | 63 ++++++++++++++++++-
 1 file changed, 62 insertions(+), 1 deletion(-)

diff --git a/litellm/llms/firecrawl/fetch/transformation.py b/litellm/llms/firecrawl/fetch/transformation.py
index 3c82ffbde3e..98714374403 100644
--- a/litellm/llms/firecrawl/fetch/transformation.py
+++ b/litellm/llms/firecrawl/fetch/transformation.py
@@ -16,11 +16,16 @@ from litellm.secret_managers.main import get_secret_str
 class FirecrawlFetchConfig(BaseFetchConfig):
     """
     Firecrawl fetch configuration.
-    Uses the /scrape endpoint to fetch and convert web content to markdown.
+    Uses the /scrape endpoint to fetch and scrape web content.
+    Supports both Firecrawl Cloud and self-hosted instances.
     """
 
     FIRECRAWL_API_BASE = "https://api.firecrawl.dev/v1"
 
+    def __init__(self) -> None:
+        super().__init__()
+        self._api_base: Optional[str] = None
+
     @staticmethod
     def ui_friendly_name() -> str:
         return "Firecrawl"
@@ -42,6 +47,62 @@ class FirecrawlFetchConfig(BaseFetchConfig):
             )
         headers["Authorization"] = f"Bearer {api_key}"
         headers["Content-Type"] = "application/json"
+
+        # Store api_base for later use in get_complete_url
+        if api_base:
+            self._api_base = api_base
+
+        return headers
+
+    def get_complete_url(self, api_base: Optional[str] = None) -> str:
+        """Get complete URL for Fetch endpoint."""
+        # Priority: explicit argument > stored > env var > default
+        effective_api_base = (
+            api_base
+            or self._api_base
+            or get_secret_str("FIRECRAWL_API_BASE")
+            or self.FIRECRAWL_API_BASE
+        )
+
+        # Append "/scrape" to the api base if it's not already there
+        if not effective_api_base.endswith("/scrape"):
+            effective_api_base = f"{effective_api_base}/scrape"
+
+        return effective_api_base
+
+    async def afetch_url(
+        self,
+        url: str,
+        headers: Dict[str, str],
+        optional_params: Dict[str, Any],
+    ) -> WebFetchResponse:
+        """
+        Fetch content from a URL using Firecrawl's /scrape endpoint.
+
+        Supports self-hosted Firecrawl via api_base configuration.
+        """
+        request_data = {
+            "url": url,
+            "formats": optional_params.get("formats", ["markdown"]),
+            "onlyMainContent": optional_params.get("onlyMainContent", True),
+        }
+
+        # Add optional parameters
+        for key in ["includeTags", "excludeTags", "headers", "waitFor", "timeout"]:
+            if key in optional_params:
+                request_data[key] = optional_params[key]
+
+        # Allow passing api_base via optional_params for dynamic override
+        api_base = optional_params.get("api_base")
+
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            response = await client.post(
+                self.get_complete_url(api_base=api_base),
+                headers=headers,
+                json=request_data,
+            )
+        headers["Authorization"] = f"Bearer {api_key}"
+        headers["Content-Type"] = "application/json"
         return headers
 
     def get_complete_url(self, api_base: Optional[str] = None) -> str: