converse-mcp-server 2.29.1 → 3.0.1

package/.env.example

@@ -43,10 +43,13 @@ MISTRAL_API_KEY=your_mistral_api_key_here
 DEEPSEEK_API_KEY=your_deepseek_api_key_here
 
 # Get your OpenRouter API key from: https://openrouter.ai/keys
+# Any provider/model slug works directly (e.g. z-ai/glm-5.2, anthropic/claude-sonnet-5);
+# append :online to a slug to opt into web search (adds a real per-request cost).
 OPENROUTER_API_KEY=your_openrouter_api_key_here
 
-# OpenRouter requires a referer for compliance (your app URL or GitHub repo)
-OPENROUTER_REFERER=https://github.com/FallDownTheSystem/converse
+# Optional: referer and title for OpenRouter ranking credit (both optional; omitting is valid)
+# OPENROUTER_REFERER=https://github.com/FallDownTheSystem/converse
+# OPENROUTER_TITLE=Converse
 
 # ============================================
 # Codex Configuration (Optional)
@@ -86,6 +89,6 @@ OPENROUTER_REFERER=https://github.com/FallDownTheSystem/converse
 # ============================================
 
 # Disable async execution tools (check_status, cancel_job) and async parameter
-# When true, removes async capabilities from chat/consensus tools
+# When true, removes async capabilities from the chat tool
 # DISABLE_ASYNC_TOOLS=false
 

package/README.md

@@ -44,7 +44,7 @@ claude mcp add converse \
   -e DEEPSEEK_API_KEY=your_key_here \
   -e OPENROUTER_API_KEY=your_key_here \
   -e ENABLE_RESPONSE_SUMMARIZATION=true \
-  -e SUMMARIZATION_MODEL=gpt-5 \
+  -e SUMMARIZATION_MODEL=gpt-5-nano \
   -s user \
   npx converse-mcp-server
 ```
@@ -68,7 +68,7 @@ Add this configuration to your Claude Desktop settings:
         "DEEPSEEK_API_KEY": "your_key_here",
         "OPENROUTER_API_KEY": "your_key_here",
         "ENABLE_RESPONSE_SUMMARIZATION": "true",
-        "SUMMARIZATION_MODEL": "gpt-5"
+        "SUMMARIZATION_MODEL": "gpt-5-nano"
       }
     }
   }
@@ -83,7 +83,7 @@ Add this configuration to your Claude Desktop settings:
   "args": ["/c", "npx", "converse-mcp-server"],
   "env": {
     "ENABLE_RESPONSE_SUMMARIZATION": "true",
-    "SUMMARIZATION_MODEL": "gpt-5"
+    "SUMMARIZATION_MODEL": "gpt-5-nano"
     // ... add your API keys here
   }
 }
@@ -94,7 +94,7 @@ Add this configuration to your Claude Desktop settings:
 Once installed, you can:
 
 - **Chat with a specific model**: Ask Claude to use the chat tool with your preferred model
-- **Get consensus**: Ask Claude to use the consensus tool when you need multiple perspectives
+- **Get consensus**: Ask Claude to use the chat tool with `mode: "consensus"` when you need multiple perspectives
 - **Run tasks in background**: Use `async: true` for long-running operations that you can check later
 - **Monitor progress**: Use the check_status tool to monitor async operations with AI-generated summaries
 - **Cancel jobs**: Use the cancel_job tool to stop running operations
@@ -105,70 +105,52 @@ Once installed, you can:
 
 ### 1. Chat Tool
 
-Talk to any AI model with support for files, images, and conversation history. The tool automatically routes your request to the right provider based on the model name. When AI summarization is enabled, generates smart titles and summaries for better context understanding.
+One tool, three modes. Pass a `models` array and choose a `mode`. Supports files, images, conversation history, and background execution. The tool routes each model to the right provider by name; `"auto"` picks the first available provider. When AI summarization is enabled, it generates smart titles and summaries.
 
 ```javascript
-// Synchronous execution (default)
+// mode "chat" (default) — 1..N models answer independently, in parallel
 {
   "prompt": "How should I structure the authentication module for this Express.js API?",
-  "model": "gemini-2.5-flash",         // Routes to Google
+  "models": ["gemini-2.5-flash"],      // Routes to Google
   "files": ["/path/to/src/auth.js", "/path/to/config.json"],
   "images": ["/path/to/architecture.png"],
-  "temperature": 0.5,
-  "reasoning_effort": "medium",
-  "use_websearch": false
+  "reasoning_effort": "medium"
 }
 
-// Asynchronous execution (for long-running tasks)
+// mode "consensus" — ≥2 models answer, then refine after seeing each other
 {
-  "prompt": "Analyze this large codebase and provide optimization recommendations",
-  "model": "gpt-5",
-  "files": ["/path/to/large-project"],
-  "async": true,           // Enables background processing
-  "continuation_id": "my-analysis-task"  // Optional: custom ID for tracking
+  "prompt": "Should we use microservices or a monolith for our e-commerce platform?",
+  "models": ["gpt-5.6", "gemini-2.5-flash", "grok-4.5"],
+  "mode": "consensus",
+  "files": ["/path/to/requirements.md"]
 }
 
-// Codex - Agentic coding assistant with local file access
+// mode "roundtable" — models speak SEQUENTIALLY in the given order, each seeing
+// the running transcript. One call = one lap; pass continuation_id for more laps.
 {
-  "prompt": "Analyze this codebase and suggest improvements",
-  "model": "codex",
-  "files": ["/path/to/your/project"],
-  "async": true  // Recommended for Codex (responses take 6-20+ seconds)
+  "prompt": "Critique this caching strategy and propose improvements.",
+  "models": ["codex", "gemini", "claude"],  // ORDER MATTERS
+  "mode": "roundtable"
+}
+
+// Asynchronous execution (for long-running tasks) — any mode
+{
+  "prompt": "Analyze this large codebase and provide optimization recommendations",
+  "models": ["gpt-5.6"],
+  "files": ["/path/to/large-project"],
+  "async": true,                         // Enables background processing
+  "continuation_id": "my-analysis-task"  // Optional: custom ID for tracking
 }
 ```
 
 **Codex Notes:**
 
-- Uses thread-based sessions (context persists with `continuation_id`)
+- Uses thread-based sessions in `chat` mode (context persists with `continuation_id`)
 - Responses typically take 6-20 seconds (complex tasks may take minutes)
 - Accesses files directly from your working directory
 - Configure sandbox mode via `CODEX_SANDBOX_MODE` environment variable
 
-### 2. Consensus Tool
-
-Get multiple AI models to analyze the same question simultaneously. Each model can see and respond to the others' answers, creating a rich discussion.
-
-```javascript
-// Synchronous consensus (default)
-{
-  "prompt": "Should we use microservices or monolith architecture for our e-commerce platform?",
-  "models": ["gpt-5", "gemini-2.5-flash", "grok-4"],
-  "files": ["/path/to/requirements.md"],
-  "enable_cross_feedback": true,
-  "temperature": 0.2
-}
-
-// Asynchronous consensus (for complex analysis)
-{
-  "prompt": "Review our system architecture and provide comprehensive recommendations",
-  "models": ["gpt-5", "gemini-2.5-pro", "claude-sonnet-4-6"],
-  "files": ["/path/to/architecture-docs"],
-  "async": true,          // Run in background
-  "enable_cross_feedback": true
-}
-```
-
-### 3. Check Status Tool
+### 2. Check Status Tool
 
 Monitor the progress and retrieve results from asynchronous operations. When AI summarization is enabled, provides intelligent summaries of ongoing and completed tasks.
 
@@ -189,7 +171,7 @@ Monitor the progress and retrieve results from asynchronous operations. When AI
 }
 ```
 
-### 4. Cancel Job Tool
+### 3. Cancel Job Tool
 
 Cancel running asynchronous operations when needed.
 
@@ -229,19 +211,16 @@ SUMMARIZATION_MODEL=gpt-5-nano        # Default: gpt-5-nano
 
 ### OpenAI Models
 
-- **gpt-5**: Latest flagship model (400K context, 128K output) - Superior reasoning, code generation, and analysis
-- **gpt-5-mini**: Faster, cost-efficient GPT-5 (400K context, 128K output) - Well-defined tasks, precise prompts
-- **gpt-5-nano**: Fastest, most cost-efficient GPT-5 (400K context, 128K output) - Summarization, classification
-- **gpt-5-pro**: Most advanced reasoning model (400K context, 272K output) - Hardest problems, extended compute time (EXPENSIVE)
-- **o3**: Strong reasoning (200K context)
-- **o3-mini**: Fast O3 variant (200K context)
-- **o3-pro**: Professional-grade reasoning (200K context) - EXTREMELY EXPENSIVE
-- **o3-deep-research**: Deep research model (200K context) - 30-90 min runtime
-- **o4-mini**: Latest reasoning model (200K context)
-- **o4-mini-deep-research**: Fast deep research model (200K context) - 15-60 min runtime
-- **gpt-4.1**: Advanced reasoning (1M context)
-- **gpt-4o**: Multimodal flagship (128K context)
-- **gpt-4o-mini**: Fast multimodal (128K context)
+- **gpt-5.6-sol** (default; aliases: `gpt-5.6`, `gpt-5`, `sol`): Flagship GPT-5.6 (1M context, 128K output) - Frontier reasoning, coding, and agentic workflows
+- **gpt-5.6-terra** (alias: `terra`): Lower-cost GPT-5.6 (400K context, 128K output) - Performance competitive with the flagship at half the price
+- **gpt-5.6-luna** (alias: `luna`): Fastest, most affordable GPT-5.6 (400K context, 128K output) - High-volume, latency-sensitive workloads
+- **gpt-5.4**: Flagship-class reasoning (1M context, 128K output)
+- **gpt-5.4-pro** (alias: `gpt-5-pro`): Maximum-performance reasoning (1M context, 272K output) - Hardest problems, extended compute time (EXPENSIVE)
+- **gpt-5-mini**, **gpt-5-nano**: Faster, cost-efficient GPT-5 tiers (400K context, 128K output)
+- **gpt-5.4-mini**, **gpt-5.4-nano**: Fast, efficient GPT-5.4 tiers (400K context, 128K output)
+- **o3**, **o3-pro**, **o4-mini**: Advanced reasoning models (200K context)
+- **gpt-4.1**: Large context (1M tokens, 32K output)
+- **o3-deep-research** (30-90 min runtime), **o4-mini-deep-research** (15-60 min runtime): Deep research models (200K context)
 
 ### Google/Gemini Models
 
@@ -253,18 +232,17 @@ SUMMARIZATION_MODEL=gpt-5-nano        # Default: gpt-5-nano
 
 **Supported Models**:
 
-- **gemini-3-pro-preview** (aliases: `pro`, `gemini`): Enhanced reasoning with thinking levels (1M context, 64K output)
-- **gemini-2.5-flash** (alias: `flash`): Ultra-fast (1M context, 65K output)
+- **gemini-3.1-pro-preview** (aliases: `pro`, `gemini-pro`): Most advanced reasoning with expanded thinking levels (1M context, 64K output)
+- **gemini-3.5-flash** (aliases: `gemini-3.5`, `flash-3.5`): Frontier-level agentic and coding performance at Flash speed (1M context, 65K output)
 - **gemini-2.5-pro** (alias: `pro 2.5`): Deep reasoning with thinking budget (1M context, 65K output)
-- **gemini-2.0-flash**: Latest with experimental thinking (1M context, 65K output)
-- **gemini-2.0-flash-lite**: Lightweight fast model, text-only (1M context, 65K output)
+- **gemini-2.5-flash** (alias: `flash`): Ultra-fast (1M context, 65K output)
+- **gemini-2.5-flash-lite** (alias: `flash-lite`): Lightweight fast model (1M context, 65K output)
 
-**Note**: Default aliases (`gemini`, `pro`) now point to Gemini 3.0 Pro. Use `gemini-2.5-pro` explicitly if you need version 2.5.
+**Note**: The bare aliases `pro` and `gemini-pro` route to Gemini 3.1 Pro through the Google API. The short name `gemini` (and `gemini:pro`/`gemini:flash`) routes to the Antigravity CLI provider instead — see below.
 
 ### X.AI/Grok Models
 
-- **grok-4-0709** (aliases: `grok`, `grok-4`): Latest advanced model (256K context)
-- **grok-code-fast-1**: Speedy and economical reasoning model that excels at agentic coding (256K context)
+- **grok-4.5** (default; aliases: `grok`, `grok-4.5-latest`, `grok-build-latest`): Flagship model with image input, reasoning content, and native web/X search (500K context). Reasoning maps to `low`/`medium`/`high` and cannot be disabled; web search is automatic. Older Grok IDs still pass through as explicit model strings.
 
 ### Anthropic Models
 
@@ -277,20 +255,28 @@ SUMMARIZATION_MODEL=gpt-5-nano        # Default: gpt-5-nano
 
 ### Mistral Models
 
-- **magistral-medium**: Frontier-class reasoning model (40K context)
-- **magistral-small**: Small reasoning model (40K context)
-- **mistral-medium-3**: Frontier-class multimodal model (128K context)
+- **mistral-medium-3-5** (default; aliases: `mistral`, `mistral-medium`): Frontier-class multimodal model with adjustable reasoning (256K context)
+- **mistral-small-2603** (alias: `mistral-small`): Hybrid multimodal model unifying instruct, reasoning, and coding (256K context)
+- **mistral-large-2512** (alias: `mistral-large`): Open-weight MoE flagship, image-capable, no adjustable reasoning (256K context)
+
+Reasoning maps to `high` (enabled) or `none` (disabled) on Medium 3.5 and Small; Large has no adjustable reasoning.
 
 ### DeepSeek Models
 
-- **deepseek-chat**: Strong MoE model with 671B/37B parameters (64K context)
-- **deepseek-reasoner**: Advanced reasoning model with CoT (64K context)
+- **deepseek-v4-pro** (default; aliases: `deepseek`, `deepseek-pro`): Flagship MoE model with thinking mode (1M context, 384K max output, text-only)
+- **deepseek-v4-flash** (alias: `deepseek-flash`): Faster, lower-cost V4 tier with thinking mode (1M context, 384K max output, text-only)
+
+Thinking mode maps `reasoning_effort` to `none` (off), `high` (enabled levels), or `max`.
 
 ### OpenRouter Models
 
-- **qwen3-235b-thinking**: Qwen3 with enhanced reasoning (32K context)
-- **qwen3-coder**: Specialized for programming tasks (32K context)
-- **kimi-k2**: Moonshot AI Kimi K2 with extended context (200K context)
+- **z-ai/glm-5.2** (default; aliases: `glm`, `glm-5.2`): Large-scale reasoning model, text-only (1M context)
+- **deepseek/deepseek-v4-pro**, **deepseek/deepseek-v4-flash**: DeepSeek V4 reasoning tiers, text-only (1M context)
+- **qwen/qwen3.7-max**, **qwen/qwen3.7-plus**: Flagship Qwen tiers (1M context; `plus` is image-capable)
+- **moonshotai/kimi-k2.7-code**, **moonshotai/kimi-k2.6**: Image-capable Moonshot models (256K context; `k2.7-code` always reasons)
+- **openrouter/auto**: Auto-selects the best model for your prompt
+
+Any other model works via its full `provider/model` slug or the `openrouter:` namespace — no extra configuration. Append `:online` to a slug to opt into web search (adds a real per-request cost).
 
 ### Codex Models
 
@@ -309,6 +295,15 @@ SUMMARIZATION_MODEL=gpt-5-nano        # Default: gpt-5-nano
   - Direct filesystem access from working directory
   - Unknown `claude:`-prefixed names pass through to the SDK (e.g. `claude:claude-sonnet-4-6`)
 
+### GitHub Copilot SDK Models
+
+Reach these with the `copilot:` namespace (e.g. `copilot:gpt-5.6-terra`); uses your GitHub Copilot subscription (`gh auth login`) - no API key needed:
+
+- **OpenAI**: `gpt-5.6-sol` (aliases: `gpt-5.6`, `gpt-5`), `gpt-5.6-terra`, `gpt-5.6-luna` (all support `reasoning_effort`)
+- **Anthropic**: `claude-fable-5` (alias: `fable`), `claude-sonnet-5` (alias: `sonnet`), `claude-opus-4.8` (aliases: `opus`, `claude`)
+- **Google**: `gemini-3.1-pro-preview` (aliases: `gemini`, `gemini-3.1-pro`), `gemini-3.5-flash` (alias: `gemini-flash`)
+- Any other `copilot:<id>` is forwarded to the Copilot backend verbatim
+
 ## 📚 Help & Documentation
 
 ### Built-in Help
@@ -352,10 +347,9 @@ LOG_LEVEL=info
 ENABLE_RESPONSE_SUMMARIZATION=true    # Enable AI-generated titles and summaries
 SUMMARIZATION_MODEL=gpt-5-nano        # Model to use for summarization (default: gpt-5-nano)
 
-# Optional: OpenRouter configuration
+# Optional: OpenRouter attribution (for ranking credit; both optional)
 OPENROUTER_REFERER=https://github.com/FallDownTheSystem/converse
 OPENROUTER_TITLE=Converse
-OPENROUTER_DYNAMIC_MODELS=true
 
 # Optional: Codex configuration
 CODEX_API_KEY=your_codex_api_key_here       # Optional if ChatGPT login available
@@ -399,27 +393,30 @@ Use `"auto"` for automatic model selection, or specify exact models:
 
 // Specific models
 "gemini-2.5-flash";
-"gpt-5";
-"grok-4-0709";
+"gpt-5.6";
+"grok-4.5";
+"z-ai/glm-5.2"; // -> OpenRouter (full slug)
+"z-ai/glm-5.2:online"; // -> OpenRouter with web search opt-in
 
 // Using aliases
 "flash"; // -> gemini-2.5-flash
-"pro"; // -> gemini-2.5-pro
-"grok"; // -> grok-4-0709
-"grok-4"; // -> grok-4-0709
+"pro"; // -> gemini-3.1-pro-preview
+"grok"; // -> grok-4.5
+"deepseek"; // -> deepseek-v4-pro
+"mistral"; // -> mistral-medium-3-5
 "fable"; // -> claude-fable-5 (Anthropic API)
 "opus"; // -> claude-opus-4-8 (Anthropic API)
 
 // SDK providers (subscription-based, no API key)
 "claude"; // -> Claude Agent SDK (Claude Fable 5)
 "claude:opus"; // -> Claude Agent SDK (Claude Opus 4.8)
-"copilot:codex"; // -> GitHub Copilot SDK (gpt-5.3-codex)
+"copilot:gpt-5.6-terra"; // -> GitHub Copilot SDK
 ```
 
 **Auto Model Behavior:**
 
-- **Chat Tool**: Selects the first available provider and uses its default model
-- **Consensus Tool**: When using `["auto"]`, automatically expands to the first 3 available providers
+- **chat mode**: `["auto"]` selects the first available provider and uses its default model
+- **consensus mode**: `["auto"]` automatically expands to the first 3 available providers
 
 Provider priority order (subscription-based SDK providers first, then API-key providers):
 
@@ -427,13 +424,13 @@ Provider priority order (subscription-based SDK providers first, then API-key pr
 2. Gemini via Antigravity CLI (`gemini`, `gemini:flash`)
 3. Claude Agent SDK (`claude` → Claude Fable 5)
 4. Copilot (`copilot`)
-5. OpenAI (`gpt-5`)
+5. OpenAI (`gpt-5.6`)
 6. Google (`gemini-pro`)
-7. XAI (`grok-4-0709`)
+7. XAI (`grok-4.5`)
 8. Anthropic (`claude-sonnet-4-20250514`)
-9. Mistral (`magistral-medium-2506`)
-10. DeepSeek (`deepseek-reasoner`)
-11. OpenRouter (`qwen/qwen3-coder`)
+9. Mistral (`mistral-medium-3-5`)
+10. DeepSeek (`deepseek-v4-pro`)
+11. OpenRouter (`z-ai/glm-5.2`)
 
 The system will use the first 3 providers that are available (authenticated SDK or valid API key). This enables automatic multi-model consensus without manually specifying models.
 
@@ -668,11 +665,16 @@ converse/
 │   │   ├── mistral.js        # Mistral AI provider
 │   │   ├── deepseek.js       # DeepSeek provider
 │   │   ├── openrouter.js     # OpenRouter provider
-│   │   └── openai-compatible.js # Base for OpenAI-compatible APIs
+│   │   ├── openrouter-discovery.js # Request-local OpenRouter slug discovery
+│   │   ├── openai-compatible.js # Base for OpenAI-compatible APIs
+│   │   ├── codex.js          # Codex agentic SDK provider
+│   │   ├── claude.js         # Claude Agent SDK provider
+│   │   ├── gemini-cli.js     # Gemini via Antigravity CLI provider
+│   │   └── copilot.js        # GitHub Copilot SDK provider
 │   ├── tools/                # MCP tool implementations
 │   │   ├── index.js          # Tool registry
-│   │   ├── chat.js           # Chat tool
-│   │   └── consensus.js      # Consensus tool
+│   │   ├── chat.js           # Unified chat tool (chat/consensus/roundtable modes)
+│   │   └── modes/            # parallel.js + roundtable.js execution engines
 │   └── utils/                # Utility modules
 │       ├── contextProcessor.js # File/image processing
 │       ├── errorHandler.js   # Error handling