feat(New Model): Add support for gpt5

config/agent_gaia-validation-gpt5.yaml

@@ -0,0 +1,79 @@
+defaults:
+  - benchmark: gaia-validation
+  - override hydra/job_logging: none
+  - _self_  # Allow defining variables at the top of this file
+
+
+main_agent:
+  prompt_class: MainAgentPrompt_GAIA
+  llm: 
+    provider_class: "GPT5OpenAIClient"
+    model_name: "gpt-5"
+    async_client: true
+    temperature: 1.0
+    top_p: 1.0
+    min_p: 0.0
+    top_k: -1
+    max_tokens: 128000
+    reasoning_effort: "high"
+    openai_api_key: "${oc.env:OPENAI_API_KEY,???}"
+    openai_base_url: "${oc.env:OPENAI_BASE_URL,https://api.openai.com/v1}"
+    openrouter_provider: ""
+    disable_cache_control: true
+    keep_tool_result: -1
+    oai_tool_thinking: false
+
+  tool_config:
+    - tool-reasoning
+
+  max_turns: -1  # Maximum number of turns for main agent execution
+  max_tool_calls_per_turn: 10  # Maximum number of tool calls per turn
+
+  input_process:
+    hint_generation: true
+    hint_llm_base_url: "${oc.env:HINT_LLM_BASE_URL,https://api.openai.com/v1}"
+  output_process:
+    final_answer_extraction: true
+    final_answer_llm_base_url: "${oc.env:FINAL_ANSWER_LLM_BASE_URL,https://api.openai.com/v1}"
+
+  openai_api_key: "${oc.env:OPENAI_API_KEY,???}" # used for hint generation and final answer extraction
+  add_message_id: true
+  keep_tool_result: -1
+  chinese_context: "${oc.env:CHINESE_CONTEXT,false}"
+
+
+sub_agents:
+  agent-worker:
+    prompt_class: SubAgentWorkerPrompt
+    llm: 
+      provider_class: "GPT5OpenAIClient"
+      model_name: "gpt-5"
+      async_client: true
+      temperature: 1.0
+      top_p: 1.0
+      min_p: 0.0
+      top_k: -1
+      max_tokens: 128000
+      reasoning_effort: "medium"
+      openai_api_key: "${oc.env:OPENAI_API_KEY,???}"
+      openai_base_url: "${oc.env:OPENAI_BASE_URL,https://api.openai.com/v1}"
+      openrouter_provider: ""
+      disable_cache_control: true
+      keep_tool_result: -1
+      oai_tool_thinking: false
+
+    tool_config:
+      - tool-searching
+      - tool-image-video
+      - tool-reading
+      - tool-code
+      - tool-audio
+
+    max_turns: -1  # Maximum number of turns for main agent execution
+    max_tool_calls_per_turn: 10  # Maximum number of tool calls per turn
+
+
+# Can define some top-level or default parameters here
+output_dir: logs/
+data_dir: "${oc.env:DATA_DIR,data}"  # Points to where data is stored
+

config/tool/tool-reading.yaml

@@ -2,4 +2,7 @@ name: "tool-reading"
 tool_command: "python"
 args:
   - "-m"
-  - "src.tool.mcp_servers.reading_mcp_server"
+  - "src.tool.mcp_servers.reading_mcp_server"
+env:
+  SERPER_API_KEY: "${oc.env:SERPER_API_KEY}"
+  JINA_API_KEY: "${oc.env:JINA_API_KEY}"

docs/mkdocs/docs/gaia_validation_gpt5.md

@@ -0,0 +1,56 @@
+# GAIA Validation - GPT5
+
+MiroFlow now supports GPT-5 with MCP tool invocation, providing a unified workflow for multi-step reasoning, information integration, and scalable tool coordination.
+
+!!! info "Prerequisites"
+    Before proceeding, please review the [GAIA Validation Prerequisites](gaia_validation_prerequisites.md) document, which covers common setup requirements, dataset preparation, and API key configuration.
+
+---
+
+## Running the Evaluation
+
+### Step 1: Dataset Preparation
+
+Follow the [dataset preparation instructions](gaia_validation_prerequisites.md#dataset-preparation) in the prerequisites document.
+
+### Step 2: API Keys Configuration
+
+Configure the following API keys in your `.env` file:
+
+```env title="GPT-5 .env Configuration"
+# Search and web scraping capabilities
+SERPER_API_KEY="your-serper-api-key"
+JINA_API_KEY="your-jina-api-key"
+
+# Code execution environment
+E2B_API_KEY="your-e2b-api-key"
+
+# Vision understanding capabilities
+ANTHROPIC_API_KEY="your-anthropic-api-key"
+GEMINI_API_KEY="your-gemini-api-key"
+
+# Primary LLM provider, LLM judge, reasoning, and hint generation
+OPENAI_API_KEY="your-openai-api-key"
+OPENAI_BASE_URL="https://api.openai.com/v1"
+
+```
+
+### Step 3: Run the Evaluation
+
+Execute the evaluation using the GPT-5 configuration:
+
+```bash title="Run GAIA Validation with GPT-5"
+uv run main.py common-benchmark \
+  --config_file_name=agent_gaia-validation-gpt5 \
+  output_dir="logs/gaia-validation-gpt5/$(date +"%Y%m%d_%H%M")"
+```
+
+### Step 4: Monitor Progress
+
+Follow the [progress monitoring instructions](gaia_validation_prerequisites.md#progress-monitoring-and-resume) in the prerequisites document.
+
+
+---
+
+!!! info "Documentation Info"
+    **Last Updated:** October 2025 · **Doc Contributor:** Team @ MiroMind AI

docs/mkdocs/docs/llm_clients_overview.md

@@ -9,6 +9,7 @@ MiroFlow supports multiple LLM providers through a unified client interface. Eac
 | `ClaudeAnthropicClient` | Anthropic Direct | claude-3-7-sonnet | `ANTHROPIC_API_KEY`, `ANTHROPIC_BASE_URL` |
 | `ClaudeOpenRouterClient` | OpenRouter | anthropic/claude-3.7-sonnet, and other [supported models](https://openrouter.ai/models) | `OPENROUTER_API_KEY`, `OPENROUTER_BASE_URL` |
 | `GPTOpenAIClient` | OpenAI | gpt-4, gpt-3.5 | `OPENAI_API_KEY`, `OPENAI_BASE_URL` |
+| `GPT5OpenAIClient` | OpenAI | gpt-5 | `OPENAI_API_KEY`, `OPENAI_BASE_URL` |
 | `MiroThinkerSGLangClient` | SGLang | MiroThinker series | `OAI_MIROTHINKER_API_KEY`, `OAI_MIROTHINKER_BASE_URL` |
 
 ## Basic Configuration
@@ -31,4 +32,4 @@ main_agent:
 ---
 
 !!! info "Documentation Info"
-    **Last Updated:** September 2025 · **Doc Contributor:** Team @ MiroMind AI
+    **Last Updated:** October 2025 · **Doc Contributor:** Team @ MiroMind AI

docs/mkdocs/docs/openai-gpt.md

@@ -1,8 +1,45 @@
 # OpenAI GPT Models
 
-OpenAI's latest models including GPT-4o and advanced reasoning models with strong coding, vision, and reasoning capabilities.
+OpenAI's latest models including GPT-5, GPT-4o and advanced reasoning models with strong coding, vision, and reasoning capabilities.
 
-## Client Used
+## Client Used for GPT-5
+
+`GPT5OpenAIClient`
+
+## Environment Setup
+
+```bash title="Environment Variables"
+export OPENAI_API_KEY="your-openai-key"
+export OPENAI_BASE_URL="https://api.openai.com/v1"  # optional
+```
+
+## Configuration
+
+```yaml title="Agent Configuration"
+main_agent:
+  llm: 
+    provider_class: "GPT5OpenAIClient"
+    model_name: "gpt-5"
+    async_client: true
+    temperature: 1.0
+    top_p: 1.0
+    min_p: 0.0
+    top_k: -1
+    max_tokens: 128000
+    reasoning_effort: "high" # Use high in the main agent, and use the default medium in the sub-agent.
+    openai_api_key: "${oc.env:OPENAI_API_KEY,???}"
+    openai_base_url: "${oc.env:OPENAI_BASE_URL,https://api.openai.com/v1}"
+```
+
+## Usage
+
+```bash title="Example Command"
+# Create custom OpenAI config
+uv run main.py trace --config_file_name=your_config_file \
+    --task="Your task" --task_file_name="data/file.txt"
+```
+
+## Client Used for GPT-4o
 
 `GPTOpenAIClient`
 
@@ -32,7 +69,10 @@ uv run main.py trace --config_file_name=your_config_file \
     --task="Your task" --task_file_name="data/file.txt"
 ```
 
+!!! note "Configuration Notes"
+    - `GPTOpenAIClient` also supports GPT-5, but it has not been fully validated on MiroFlow yet. We recommend using `GPT5OpenAIClient`.
+
 ---
 
 !!! info "Documentation Info"
-    **Last Updated:** September 2025 · **Doc Contributor:** Team @ MiroMind AI
+    **Last Updated:** October 2025 · **Doc Contributor:** Team @ MiroMind AI

docs/mkdocs/mkdocs.yml

@@ -52,6 +52,7 @@ nav:
       - GAIA-Validation:
         - Prerequisites: gaia_validation_prerequisites.md
         - Claude-3.7-Sonnet: gaia_validation_claude37sonnet.md
+        - GPT-5: gaia_validation_gpt5.md
         - MiroThinker: gaia_validation_mirothinker.md
       - GAIA-Validation-Text-Only: gaia_validation_text_only.md
       - GAIA-Test: gaia_test.md

src/llm/provider_client_base.py

@@ -43,6 +43,7 @@ def __post_init__(self):
         self.top_p: float = self.cfg.llm.top_p
         self.min_p: float = self.cfg.llm.min_p
         self.top_k: int = self.cfg.llm.top_k
+        self.reasoning_effort: str = self.cfg.llm.get("reasoning_effort", "medium")
         self.repetition_penalty: float = self.cfg.llm.get("repetition_penalty", 1.0)
         self.max_tokens: int = self.cfg.llm.max_tokens
         self.max_context_length: int = self.cfg.llm.get("max_context_length", -1)

src/llm/providers/claude_openrouter_client.py

@@ -411,4 +411,4 @@ def _apply_cache_control(self, messages):
             else:
                 # Other messages add directly
                 cached_messages.append(turn)
-        return list(reversed(cached_messages))
+        return list(reversed(cached_messages))