re-init

.gitignore

@@ -46,6 +46,12 @@ wheels/
 # misc
 .DS_Store
 *.pem
+*.save
+.promptfoo/
+.promptfoo-home/
+.tmp/
+promptfoo.db
+data-gym-cache/
 
 # debug
 npm-debug.log*

.vscode/tasks.json

@@ -0,0 +1,141 @@
+{
+  "version": "2.0.0",
+  "tasks": [
+    {
+      "label": "DeepWiki: Install API",
+      "type": "shell",
+      "command": "poetry",
+      "args": [
+        "-C",
+        "${workspaceFolder}/api",
+        "install"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: API",
+      "type": "shell",
+      "command": "${workspaceFolder}/api/.venv/bin/python",
+      "args": [
+        "-m",
+        "api.main"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}",
+        "env": {
+          "PYTHONPATH": "${workspaceFolder}"
+        }
+      },
+      "isBackground": true,
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: Web",
+      "type": "shell",
+      "command": "env",
+      "args": [
+        "-u",
+        "ELECTRON_RUN_AS_NODE",
+        "npm",
+        "run",
+        "dev"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "isBackground": true,
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: Start All",
+      "dependsOn": [
+        "DeepWiki: API",
+        "DeepWiki: Web"
+      ],
+      "dependsOrder": "parallel"
+    },
+    {
+      "label": "DeepWiki: Promptfoo Smoke",
+      "type": "shell",
+      "command": "npm",
+      "args": [
+        "run",
+        "promptfoo:smoke"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: Promptfoo Eval",
+      "type": "shell",
+      "command": "npm",
+      "args": [
+        "run",
+        "promptfoo:eval"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: Promptfoo View",
+      "type": "shell",
+      "command": "npm",
+      "args": [
+        "run",
+        "promptfoo:view"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    },
+    {
+      "label": "DeepWiki: Promptfoo CI",
+      "type": "shell",
+      "command": "npm",
+      "args": [
+        "run",
+        "promptfoo:ci"
+      ],
+      "options": {
+        "cwd": "${workspaceFolder}"
+      },
+      "presentation": {
+        "reveal": "always",
+        "panel": "dedicated"
+      },
+      "problemMatcher": []
+    }
+  ]
+}

README.md

@@ -122,6 +122,17 @@ yarn dev
 3. For private repositories, click "+ Add access tokens" and enter your GitHub or GitLab personal access token
 4. Click "Generate Wiki" and watch the magic happen!
 
+## Local Evals With Promptfoo
+
+DeepWiki now includes a local `promptfoo` setup for regression-style checks against your Ollama-backed API.
+
+1. Start the backend on `http://127.0.0.1:8001`
+2. Run `npm run promptfoo:smoke` for a fast local confidence check
+3. Run `npm run promptfoo:ci` for the full local regression suite
+4. Optional: inspect the latest run with `npm run promptfoo:view`
+
+The default eval config lives in `promptfooconfig.yaml` and targets the local repository path `/Users/samroku/dev/deepwiki-open`, so it stays inside your local-first workflow by default.
+
 ## 🔍 How It Works
 
 DeepWiki uses AI to:

api/api.py

@@ -180,7 +180,7 @@ async def get_model_config():
 
         # Create providers from the config file
         providers = []
-        default_provider = configs.get("default_provider", "google")
+        default_provider = configs.get("default_provider", "ollama")
 
         # Add provider configuration based on config.py
         for provider_id, provider_config in configs["providers"].items():
@@ -221,7 +221,7 @@ async def get_model_config():
                     ]
                 )
             ],
-            defaultProvider="google"
+            defaultProvider="ollama"
         )
 
 @app.post("/export/wiki")

api/config.py

@@ -48,8 +48,14 @@
 WIKI_AUTH_MODE = raw_auth_mode.lower() in ['true', '1', 't']
 WIKI_AUTH_CODE = os.environ.get('DEEPWIKI_AUTH_CODE', '')
 
+# Privacy/local-first settings
+raw_local_only_mode = os.environ.get('DEEPWIKI_LOCAL_ONLY', 'False')
+LOCAL_ONLY_MODE = raw_local_only_mode.lower() in ['true', '1', 't']
+ALLOWED_PROVIDERS_ENV = os.environ.get('DEEPWIKI_PROVIDER_ALLOWLIST', '')
+DEFAULT_PROVIDER_OVERRIDE = os.environ.get('DEEPWIKI_DEFAULT_PROVIDER', '').strip().lower()
+
 # Embedder settings
-EMBEDDER_TYPE = os.environ.get('DEEPWIKI_EMBEDDER_TYPE', 'openai').lower()
+EMBEDDER_TYPE = os.environ.get('DEEPWIKI_EMBEDDER_TYPE', 'ollama').lower()
 
 # Get configuration directory from environment variable, or use default if not set
 CONFIG_DIR = os.environ.get('DEEPWIKI_CONFIG_DIR', None)
@@ -147,6 +153,24 @@ def load_generator_config():
 
     return generator_config
 
+
+def get_allowed_providers():
+    """
+    Resolve the provider allowlist from environment variables.
+
+    When DEEPWIKI_LOCAL_ONLY is enabled and no explicit allowlist is set,
+    only the local Ollama provider is exposed.
+    """
+    if ALLOWED_PROVIDERS_ENV.strip():
+        return [
+            provider.strip().lower()
+            for provider in ALLOWED_PROVIDERS_ENV.split(",")
+            if provider.strip()
+        ]
+    if LOCAL_ONLY_MODE:
+        return ["ollama"]
+    return []
+
 # Load embedder configuration
 def load_embedder_config():
     embedder_config = load_json_config("embedder.json")
@@ -336,8 +360,22 @@ def load_lang_config():
 
 # Update configuration
 if generator_config:
-    configs["default_provider"] = generator_config.get("default_provider", "google")
-    configs["providers"] = generator_config.get("providers", {})
+    configured_providers = generator_config.get("providers", {})
+    allowed_providers = get_allowed_providers()
+
+    if allowed_providers:
+        configured_providers = {
+            provider_id: provider_config
+            for provider_id, provider_config in configured_providers.items()
+            if provider_id in allowed_providers
+        }
+
+    default_provider = DEFAULT_PROVIDER_OVERRIDE or generator_config.get("default_provider", "ollama")
+    if default_provider not in configured_providers and configured_providers:
+        default_provider = next(iter(configured_providers.keys()))
+
+    configs["default_provider"] = default_provider
+    configs["providers"] = configured_providers
 
 # Update embedder configuration
 if embedder_config:
@@ -356,7 +394,7 @@ def load_lang_config():
     configs["lang_config"] = lang_config
 
 
-def get_model_config(provider="google", model=None):
+def get_model_config(provider="ollama", model=None):
     """
     Get configuration for the specified provider and model
 

api/config/embedder.json

@@ -1,33 +1,14 @@
 {
-  "embedder": {
-    "client_class": "OpenAIClient",
-    "batch_size": 500,
-    "model_kwargs": {
-      "model": "text-embedding-3-small",
-      "dimensions": 256,
-      "encoding_format": "float"
-    }
-  },
   "embedder_ollama": {
     "client_class": "OllamaClient",
     "model_kwargs": {
       "model": "nomic-embed-text"
     }
   },
-  "embedder_google": {
-    "client_class": "GoogleEmbedderClient",
-    "batch_size": 100,
-    "model_kwargs": {
-      "model": "gemini-embedding-001",
-      "task_type": "SEMANTIC_SIMILARITY"
-    }
-  },
-  "embedder_bedrock": {
-    "client_class": "BedrockClient",
-    "batch_size": 100,
+  "embedder": {
+    "client_class": "OllamaClient",
     "model_kwargs": {
-      "model": "amazon.titan-embed-text-v2:0",
-      "dimensions": 256
+      "model": "nomic-embed-text"
     }
   },
   "retriever": {

api/config/generator.json

@@ -1,5 +1,5 @@
 {
-  "default_provider": "google",
+  "default_provider": "ollama",
   "providers": {
     "dashscope": {
       "default_model": "qwen-plus",
@@ -196,4 +196,3 @@
     }
   }
 }
-

api/data_pipeline.py

@@ -700,14 +700,35 @@ def get_file_content(repo_url: str, file_path: str, repo_type: str = None, acces
     Raises:
         ValueError: If the file cannot be fetched or if the URL is not valid
     """
+    if repo_type == "local" or (
+        repo_url
+        and not repo_url.startswith("http://")
+        and not repo_url.startswith("https://")
+    ):
+        repo_root = os.path.abspath(repo_url)
+        candidate_path = os.path.abspath(os.path.join(repo_root, file_path))
+
+        # Prevent path traversal outside the repository root.
+        if os.path.commonpath([repo_root, candidate_path]) != repo_root:
+            raise ValueError("Requested file path escapes the local repository root.")
+
+        if not os.path.isfile(candidate_path):
+            raise ValueError(f"Local file not found: {file_path}")
+
+        try:
+            with open(candidate_path, "r", encoding="utf-8") as f:
+                return f.read()
+        except Exception as e:
+            raise ValueError(f"Failed to read local file content: {str(e)}")
+
     if repo_type == "github":
         return get_github_file_content(repo_url, file_path, access_token)
     elif repo_type == "gitlab":
         return get_gitlab_file_content(repo_url, file_path, access_token)
     elif repo_type == "bitbucket":
         return get_bitbucket_file_content(repo_url, file_path, access_token)
     else:
-        raise ValueError("Unsupported repository type. Only GitHub, GitLab, and Bitbucket are supported.")
+        raise ValueError("Unsupported repository type. Only local, GitHub, GitLab, and Bitbucket are supported.")
 
 class DatabaseManager:
     """