Add configurable LLM provider layer (Gemini, SiliconFlow, OpenAI-compatible)

.env.example

@@ -1,6 +1,23 @@
-# Google Gemini API Key
+# LLM provider selection: google | siliconflow | openai
+FS_EXPLORER_LLM_PROVIDER=google
+
+# Optional model override (provider-specific default if unset)
+# FS_EXPLORER_LLM_MODEL=gemini-3-flash-preview
+# FS_EXPLORER_LLM_MODEL=Qwen/Qwen2.5-72B-Instruct
+
+# Optional base URL override for OpenAI-compatible providers
+# FS_EXPLORER_LLM_BASE_URL=https://api.siliconflow.cn/v1
+
+# --- Google Gemini ---
 # Get yours at: https://aistudio.google.com/apikey
-GOOGLE_API_KEY=your_api_key_here
+GOOGLE_API_KEY=your_google_api_key_here
+
+# --- SiliconFlow (OpenAI-compatible) ---
+# Get yours at: https://cloud.siliconflow.cn/account/ak
+SILICONFLOW_API_KEY=your_siliconflow_api_key_here
+
+# --- OpenAI (optional) ---
+# OPENAI_API_KEY=your_openai_api_key_here
 
 # Optional: dedicated key for langextract metadata mode.
 # If unset, indexing will fall back to GOOGLE_API_KEY.

PROVIDERS.md

@@ -0,0 +1,64 @@
+# LLM Provider Configuration
+
+FsExplorer supports multiple LLM backends through a small provider adapter layer.
+
+## Quick Start
+
+### Google Gemini (default)
+
+```bash
+export FS_EXPLORER_LLM_PROVIDER=google
+export GOOGLE_API_KEY=your_google_api_key
+```
+
+### SiliconFlow (OpenAI-compatible)
+
+```bash
+export FS_EXPLORER_LLM_PROVIDER=siliconflow
+export SILICONFLOW_API_KEY=your_siliconflow_api_key
+# Optional overrides
+export FS_EXPLORER_LLM_MODEL=Qwen/Qwen2.5-72B-Instruct
+export FS_EXPLORER_LLM_BASE_URL=https://api.siliconflow.cn/v1
+# International endpoint: https://api.siliconflow.com/v1
+```
+
+Get a SiliconFlow API key at https://cloud.siliconflow.cn/account/ak
+
+### OpenAI
+
+```bash
+export FS_EXPLORER_LLM_PROVIDER=openai
+export OPENAI_API_KEY=your_openai_api_key
+export FS_EXPLORER_LLM_MODEL=gpt-4o-mini
+export FS_EXPLORER_LLM_BASE_URL=https://api.openai.com/v1
+```
+
+## Environment Variables
+
+| Variable | Description |
+|----------|-------------|
+| `FS_EXPLORER_LLM_PROVIDER` | `google`, `siliconflow`, or `openai` (default: `google`) |
+| `FS_EXPLORER_LLM_MODEL` | Model id override |
+| `FS_EXPLORER_LLM_BASE_URL` | Base URL for OpenAI-compatible providers |
+| `GOOGLE_API_KEY` | Google Gemini API key |
+| `SILICONFLOW_API_KEY` | SiliconFlow API key |
+| `OPENAI_API_KEY` | OpenAI API key |
+
+## Architecture
+
+```
+FsExplorerAgent
+    -> llm.create_llm_client()
+            -> GoogleGeminiClient      (native JSON schema)
+            -> OpenAICompatibleClient  (SiliconFlow, OpenAI, ...)
+```
+
+Google Gemini uses native structured JSON output. OpenAI-compatible providers use `response_format=json_object` plus the Action JSON schema embedded in the system prompt, with flexible parsing for imperfect model output.
+
+## Embeddings / Indexing
+
+Vector indexing (`explore index --with-embeddings`) still uses Google Gemini embeddings by default via `GOOGLE_API_KEY`. Chat provider selection is independent of embedding configuration.
+
+## Security
+
+Never commit `.env` or real API keys. Use `.env.example` as a template only.

README.md

@@ -24,7 +24,7 @@ This video explains the architecture of the project and how to run it.
 
 - 🔍 **6 Tools**: `scan_folder`, `preview_file`, `parse_file`, `read`, `grep`, `glob`
 - 📄 **Document Support**: PDF, DOCX, PPTX, XLSX, HTML, Markdown (via Docling)
-- 🤖 **Powered by**: Google Gemini 3 Flash with structured JSON output
+- 🤖 **Multi-LLM**: Google Gemini, SiliconFlow, OpenAI-compatible APIs
 - 💰 **Cost Efficient**: ~$0.001 per query with token tracking
 - 🌐 **Web UI**: Real-time WebSocket streaming interface
 - 📊 **Citations**: Answers include source references
@@ -45,13 +45,19 @@ pip install .
 
 ## Configuration
 
-Create a `.env` file in the project root:
+Create a `.env` file in the project root (see `.env.example`):
 
 ```bash
+# Google Gemini (default)
+FS_EXPLORER_LLM_PROVIDER=google
 GOOGLE_API_KEY=your_api_key_here
+
+# SiliconFlow (OpenAI-compatible)
+# FS_EXPLORER_LLM_PROVIDER=siliconflow
+# SILICONFLOW_API_KEY=your_api_key_here
 ```
 
-Get your API key from [Google AI Studio](https://aistudio.google.com/apikey).
+See [PROVIDERS.md](PROVIDERS.md) for all supported backends and model overrides.
 
 ## Usage
 
@@ -125,7 +131,7 @@ uv run explore --task "Look in data/large_acquisition/. What happens to employee
 
 | Component | Technology |
 |-----------|------------|
-| LLM | Google Gemini 3 Flash |
+| LLM | Google Gemini / SiliconFlow / OpenAI-compatible |
 | Document Parsing | Docling (local, open-source) |
 | Orchestration | LlamaIndex Workflows |
 | CLI | Typer + Rich |
@@ -136,7 +142,8 @@ uv run explore --task "Look in data/large_acquisition/. What happens to employee
 
 ```
 src/fs_explorer/
-├── agent.py      # Gemini client, token tracking
+├── agent.py      # Agent + tool registry
+├── llm/          # Multi-provider LLM adapters
 ├── workflow.py   # LlamaIndex workflow engine
 ├── fs.py         # File tools: scan, parse, grep
 ├── models.py     # Pydantic models for actions

pyproject.toml

@@ -13,6 +13,7 @@ dependencies = [
     "duckdb>=1.0.0",
     "fastapi>=0.115.0",
     "google-genai>=1.55.0",
+    "openai>=1.60.0",
     "langextract>=1.0.0",
     "llama-index-workflows>=2.11.5",
     "python-dotenv>=1.0.0",