Update project files

Update README.md

README.md

@@ -1,6 +1,6 @@
-# Gemini Fullstack LangGraph Quickstart
+# OpenAI-Compatible Fullstack LangGraph Quickstart
 
-This project demonstrates a fullstack application using a React frontend and a LangGraph-powered backend agent. The agent is designed to perform comprehensive research on a user's query by dynamically generating search terms, querying the web using Google Search, reflecting on the results to identify knowledge gaps, and iteratively refining its search until it can provide a well-supported answer with citations. This application serves as an example of building research-augmented conversational AI using LangGraph and Google's Gemini models.
+This project demonstrates a fullstack application using a React frontend and a LangGraph-powered backend agent. The agent performs comprehensive research on a user's query by dynamically generating search terms, querying the web with Tavily, reflecting on the results to identify knowledge gaps, and iteratively refining its search until it can provide a well-supported answer with citations. This application serves as an example of building research-augmented conversational AI using LangGraph and an OpenAI-compatible chat model.
 
 <img src="./app.png" title="Gemini Fullstack LangGraph" alt="Gemini Fullstack LangGraph" width="90%">
 
@@ -8,11 +8,11 @@ This project demonstrates a fullstack application using a React frontend and a L
 
 - 💬 Fullstack application with a React frontend and LangGraph backend.
 - 🧠 Powered by a LangGraph agent for advanced research and conversational AI.
-- 🔍 Dynamic search query generation using Google Gemini models.
-- 🌐 Integrated web research via Google Search API.
+- 🔍 Dynamic search query generation using an OpenAI-compatible chat model.
+- 🌐 Integrated web research via Tavily Search.
 - 🤔 Reflective reasoning to identify knowledge gaps and refine searches.
 - 📄 Generates answers with citations from gathered sources.
-- 🔄 Hot-reloading for both frontend and backend development during development.
+- 🔄 Hot-reloading for both frontend and backend during development.
 
 ## Project Structure
 
@@ -29,10 +29,15 @@ Follow these steps to get the application running locally for development and te
 
 -   Node.js and npm (or yarn/pnpm)
 -   Python 3.11+
--   **`GEMINI_API_KEY`**: The backend agent requires a Google Gemini API key.
+-   **`OPENAI_API_KEY`**: The backend agent requires an OpenAI-compatible API key.
+-   **`OPENAI_BASE_URL`**: Set this to `https://open.bigmodel.cn/api/paas/v4` for BigModel.
+-   **`TAVILY_API_KEY`**: Required for web search.
     1.  Navigate to the `backend/` directory.
     2.  Create a file named `.env` by copying the `backend/.env.example` file.
-    3.  Open the `.env` file and add your Gemini API key: `GEMINI_API_KEY="YOUR_ACTUAL_API_KEY"`
+    3.  Open the `.env` file and add:
+        `OPENAI_API_KEY="YOUR_ACTUAL_API_KEY"`
+        `OPENAI_BASE_URL="https://open.bigmodel.cn/api/paas/v4"`
+        `TAVILY_API_KEY="YOUR_ACTUAL_TAVILY_API_KEY"`
 
 **2. Install Dependencies:**
 
@@ -67,11 +72,23 @@ The core of the backend is a LangGraph agent defined in `backend/src/agent/graph
 
 <img src="./agent.png" title="Agent Flow" alt="Agent Flow" width="50%">
 
-1.  **Generate Initial Queries:** Based on your input, it generates a set of initial search queries using a Gemini model.
-2.  **Web Research:** For each query, it uses the Gemini model with the Google Search API to find relevant web pages.
-3.  **Reflection & Knowledge Gap Analysis:** The agent analyzes the search results to determine if the information is sufficient or if there are knowledge gaps. It uses a Gemini model for this reflection process.
+1.  **Generate Initial Queries:** Based on your input, it generates a set of initial search queries using the configured chat model.
+2.  **Web Research:** For each query, it uses Tavily Search to gather relevant web pages and snippets.
+3.  **Reflection & Knowledge Gap Analysis:** The agent analyzes the search results to determine if the information is sufficient or if there are knowledge gaps. It uses the configured chat model for this reflection process.
 4.  **Iterative Refinement:** If gaps are found or the information is insufficient, it generates follow-up queries and repeats the web research and reflection steps (up to a configured maximum number of loops).
-5.  **Finalize Answer:** Once the research is deemed sufficient, the agent synthesizes the gathered information into a coherent answer, including citations from the web sources, using a Gemini model.
+5.  **Finalize Answer:** Once the research is deemed sufficient, the agent synthesizes the gathered information into a coherent answer, including citations from the web sources, using the configured chat model.
+
+## CLI Example
+
+For quick one-off questions you can execute the agent from the command line. The
+script `backend/examples/cli_research.py` runs the LangGraph agent and prints the
+final answer:
+
+```bash
+cd backend
+python examples/cli_research.py "What are the latest trends in renewable energy?"
+```
+
 
 ## Deployment
 
@@ -79,18 +96,18 @@ In production, the backend server serves the optimized static frontend build. La
 
 _Note: For the docker-compose.yml example you need a LangSmith API key, you can get one from [LangSmith](https://smith.langchain.com/settings)._
 
-_Note: If you are not running the docker-compose.yml example or exposing the backend server to the public internet, you update the `apiUrl` in the `frontend/src/App.tsx` file your host. Currently the `apiUrl` is set to `http://localhost:8123` for docker-compose or `http://localhost:2024` for development._
+_Note: If you are not running the docker-compose.yml example or exposing the backend server to the public internet, you should update the `apiUrl` in the `frontend/src/App.tsx` file to your host. Currently the `apiUrl` is set to `http://localhost:8123` for docker-compose or `http://localhost:2024` for development._
 
 **1. Build the Docker Image:**
 
    Run the following command from the **project root directory**:
    ```bash
-   docker build -t gemini-fullstack-langgraph -f Dockerfile .
+   docker build -t openai-fullstack-langgraph -f Dockerfile .
    ```
 **2. Run the Production Server:**
 
    ```bash
-   GEMINI_API_KEY=<your_gemini_api_key> LANGSMITH_API_KEY=<your_langsmith_api_key> docker-compose up
+   OPENAI_API_KEY=<your_api_key> OPENAI_BASE_URL=https://open.bigmodel.cn/api/paas/v4 TAVILY_API_KEY=<your_tavily_api_key> LANGSMITH_API_KEY=<your_langsmith_api_key> docker-compose up
    ```
 
 Open your browser and navigate to `http://localhost:8123/app/` to see the application. The API will be available at `http://localhost:8123`.
@@ -101,7 +118,8 @@ Open your browser and navigate to `http://localhost:8123/app/` to see the applic
 - [Tailwind CSS](https://tailwindcss.com/) - For styling.
 - [Shadcn UI](https://ui.shadcn.com/) - For components.
 - [LangGraph](https://github.com/langchain-ai/langgraph) - For building the backend research agent.
-- [Google Gemini](https://ai.google.dev/models/gemini) - LLM for query generation, reflection, and answer synthesis.
+- OpenAI-compatible chat model - LLM for query generation, reflection, and answer synthesis.
+- [Tavily](https://tavily.com/) - Web search for research retrieval.
 
 ## License
 

backend/examples/cli_research.py

@@ -0,0 +1,43 @@
+import argparse
+from langchain_core.messages import HumanMessage
+from agent.graph import graph
+
+
+def main() -> None:
+    """Run the research agent from the command line."""
+    parser = argparse.ArgumentParser(description="Run the LangGraph research agent")
+    parser.add_argument("question", help="Research question")
+    parser.add_argument(
+        "--initial-queries",
+        type=int,
+        default=3,
+        help="Number of initial search queries",
+    )
+    parser.add_argument(
+        "--max-loops",
+        type=int,
+        default=2,
+        help="Maximum number of research loops",
+    )
+    parser.add_argument(
+        "--reasoning-model",
+        default="GLM-4.5-Air",
+        help="Model for the final answer",
+    )
+    args = parser.parse_args()
+
+    state = {
+        "messages": [HumanMessage(content=args.question)],
+        "initial_search_query_count": args.initial_queries,
+        "max_research_loops": args.max_loops,
+        "reasoning_model": args.reasoning_model,
+    }
+
+    result = graph.invoke(state)
+    messages = result.get("messages", [])
+    if messages:
+        print(messages[-1].content)
+
+
+if __name__ == "__main__":
+    main()

backend/pyproject.toml

@@ -11,13 +11,13 @@ requires-python = ">=3.11,<4.0"
 dependencies = [
     "langgraph>=0.2.6",
     "langchain>=0.3.19",
-    "langchain-google-genai",
+    "langchain-openai",
     "python-dotenv>=1.0.1",
     "langgraph-sdk>=0.1.57",
     "langgraph-cli",
     "langgraph-api",
     "fastapi",
-    "google-genai",
+    "tavily-python",
 ]
 
 

backend/src/agent/configuration.py

@@ -9,21 +9,21 @@ class Configuration(BaseModel):
     """The configuration for the agent."""
 
     query_generator_model: str = Field(
-        default="gemini-2.0-flash",
+        default="GLM-4.5-Air",
         metadata={
             "description": "The name of the language model to use for the agent's query generation."
         },
     )
 
     reflection_model: str = Field(
-        default="gemini-2.5-flash",
+        default="GLM-4.5-Air",
         metadata={
             "description": "The name of the language model to use for the agent's reflection."
         },
     )
 
     answer_model: str = Field(
-        default="gemini-2.5-pro",
+        default="GLM-4.5-Air",
         metadata={
             "description": "The name of the language model to use for the agent's answer."
         },

backend/src/agent/graph.py

@@ -7,7 +7,8 @@ from langgraph.types import Send
 from langgraph.graph import StateGraph
 from langgraph.graph import START, END
 from langchain_core.runnables import RunnableConfig
-from google.genai import Client
+from langchain_openai import ChatOpenAI
+from tavily import TavilyClient
 
 from agent.state import (
     OverallState,
@@ -23,60 +24,61 @@ from agent.prompts import (
     reflection_instructions,
     answer_instructions,
 )
-from langchain_google_genai import ChatGoogleGenerativeAI
 from agent.utils import (
-    get_citations,
+    deduplicate_and_format_sources,
     get_research_topic,
-    insert_citation_markers,
-    resolve_urls,
 )
 
 load_dotenv()
 
-if os.getenv("GEMINI_API_KEY") is None:
-    raise ValueError("GEMINI_API_KEY is not set")
+if os.getenv("OPENAI_API_KEY") is None:
+    raise ValueError("OPENAI_API_KEY is not set")
 
-# Used for Google Search API
-genai_client = Client(api_key=os.getenv("GEMINI_API_KEY"))
+if os.getenv("TAVILY_API_KEY") is None:
+    raise ValueError("TAVILY_API_KEY is not set")
+
+tavily_client = TavilyClient(api_key=os.getenv("TAVILY_API_KEY"))
+
+
+def create_llm(model: str, temperature: float) -> ChatOpenAI:
+    """Create an OpenAI-compatible chat client."""
+    return ChatOpenAI(
+        model=model,
+        temperature=temperature,
+        max_retries=2,
+        api_key=os.getenv("OPENAI_API_KEY"),
+        base_url=os.getenv("OPENAI_BASE_URL", "https://open.bigmodel.cn/api/paas/v4"),
+    )
 
 
 # Nodes
 def generate_query(state: OverallState, config: RunnableConfig) -> QueryGenerationState:
-    """LangGraph node that generates a search queries based on the User's question.
+    """LangGraph node that generates search queries based on the User's question.
 
-    Uses Gemini 2.0 Flash to create an optimized search query for web research based on
-    the User's question.
+    Uses the configured chat model to create optimized search queries for web research.
 
     Args:
         state: Current graph state containing the User's question
         config: Configuration for the runnable, including LLM provider settings
 
     Returns:
-        Dictionary with state update, including search_query key containing the generated query
+        Dictionary with state update, including search_query key containing the generated queries
     """
     configurable = Configuration.from_runnable_config(config)
 
-    # check for custom initial search query count
+    # Check for custom initial search query count.
     if state.get("initial_search_query_count") is None:
         state["initial_search_query_count"] = configurable.number_of_initial_queries
 
-    # init Gemini 2.0 Flash
-    llm = ChatGoogleGenerativeAI(
-        model=configurable.query_generator_model,
-        temperature=1.0,
-        max_retries=2,
-        api_key=os.getenv("GEMINI_API_KEY"),
-    )
+    llm = create_llm(configurable.query_generator_model, temperature=1.0)
     structured_llm = llm.with_structured_output(SearchQueryList)
 
-    # Format the prompt
     current_date = get_current_date()
     formatted_prompt = query_writer_instructions.format(
         current_date=current_date,
         research_topic=get_research_topic(state["messages"]),
         number_queries=state["initial_search_query_count"],
     )
-    # Generate the search queries
     result = structured_llm.invoke(formatted_prompt)
     return {"search_query": result.query}
 
@@ -93,46 +95,34 @@ def continue_to_web_research(state: QueryGenerationState):
 
 
 def web_research(state: WebSearchState, config: RunnableConfig) -> OverallState:
-    """LangGraph node that performs web research using the native Google Search API tool.
-
-    Executes a web search using the native Google Search API tool in combination with Gemini 2.0 Flash.
-
-    Args:
-        state: Current graph state containing the search query and research loop count
-        config: Configuration for the runnable, including search API settings
-
-    Returns:
-        Dictionary with state update, including sources_gathered, research_loop_count, and web_research_results
-    """
-    # Configure
-    configurable = Configuration.from_runnable_config(config)
+    """LangGraph node that performs web research using Tavily search."""
+    Configuration.from_runnable_config(config)
     formatted_prompt = web_searcher_instructions.format(
         current_date=get_current_date(),
         research_topic=state["search_query"],
     )
+    search_response = tavily_client.search(
+        query=state["search_query"],
+        topic="general",
+        search_depth="advanced",
+        max_results=5,
+        include_answer=False,
+        include_raw_content=False,
+    )
+    sources_gathered, formatted_results = deduplicate_and_format_sources(
+        search_response.get("results", []), int(state["id"])
+    )
 
-    # Uses the google genai client as the langchain client doesn't return grounding metadata
-    response = genai_client.models.generate_content(
-        model=configurable.query_generator_model,
-        contents=formatted_prompt,
-        config={
-            "tools": [{"google_search": {}}],
-            "temperature": 0,
-        },
+    web_research_result = (
+        f"{formatted_prompt}\n\n"
+        f"Search query: {state['search_query']}\n\n"
+        f"Findings:\n{formatted_results}"
     )
-    # resolve the urls to short urls for saving tokens and time
-    resolved_urls = resolve_urls(
-        response.candidates[0].grounding_metadata.grounding_chunks, state["id"]
-    )
-    # Gets the citations and adds them to the generated text
-    citations = get_citations(response, resolved_urls)
-    modified_text = insert_citation_markers(response.text, citations)
-    sources_gathered = [item for citation in citations for item in citation["segments"]]
 
     return {
         "sources_gathered": sources_gathered,
         "search_query": [state["search_query"]],
-        "web_research_result": [modified_text],
+        "web_research_result": [web_research_result],
     }
 
 
@@ -162,13 +152,7 @@ def reflection(state: OverallState, config: RunnableConfig) -> ReflectionState:
         research_topic=get_research_topic(state["messages"]),
         summaries="\n\n---\n\n".join(state["web_research_result"]),
     )
-    # init Reasoning Model
-    llm = ChatGoogleGenerativeAI(
-        model=reasoning_model,
-        temperature=1.0,
-        max_retries=2,
-        api_key=os.getenv("GEMINI_API_KEY"),
-    )
+    llm = create_llm(reasoning_model, temperature=1.0)
     result = llm.with_structured_output(Reflection).invoke(formatted_prompt)
 
     return {
@@ -241,13 +225,7 @@ def finalize_answer(state: OverallState, config: RunnableConfig):
         summaries="\n---\n\n".join(state["web_research_result"]),
     )
 
-    # init Reasoning Model, default to Gemini 2.5 Flash
-    llm = ChatGoogleGenerativeAI(
-        model=reasoning_model,
-        temperature=0,
-        max_retries=2,
-        api_key=os.getenv("GEMINI_API_KEY"),
-    )
+    llm = create_llm(reasoning_model, temperature=0)
     result = llm.invoke(formatted_prompt)
 
     # Replace the short urls with the original urls and add all used urls to the sources_gathered

backend/src/agent/prompts.py

@@ -34,11 +34,11 @@ Topic: What revenue grew more last year apple stock or the number of people buyi
 Context: {research_topic}"""
 
 
-web_searcher_instructions = """Conduct targeted Google Searches to gather the most recent, credible information on "{research_topic}" and synthesize it into a verifiable text artifact.
+web_searcher_instructions = """Conduct targeted web research to gather the most recent, credible information on "{research_topic}" and synthesize it into a verifiable text artifact.
 
 Instructions:
 - Query should ensure that the most current information is gathered. The current date is {current_date}.
-- Conduct multiple, diverse searches to gather comprehensive information.
+- Review the returned web search results and extract the strongest, most relevant findings.
 - Consolidate key findings while meticulously tracking the source(s) for each specific piece of information.
 - The output should be a well-written summary or report based on your search findings. 
 - Only include the information found in the search results, don't make up any information.
@@ -87,7 +87,7 @@ Instructions:
 - You have access to all the information gathered from the previous steps.
 - You have access to the user's question.
 - Generate a high-quality answer to the user's question based on the provided summaries and the user's question.
-- you MUST include all the citations from the summaries in the answer correctly.
+- Include the sources you used from the Summaries in the answer correctly, using markdown links. THIS IS A MUST.
 
 User Context:
 - {research_topic}

backend/src/agent/state.py

@@ -8,8 +8,6 @@ from typing_extensions import Annotated
 
 
 import operator
-from dataclasses import dataclass, field
-from typing_extensions import Annotated
 
 
 class OverallState(TypedDict):

backend/src/agent/utils.py

@@ -1,166 +1,59 @@
-from typing import Any, Dict, List
-from langchain_core.messages import AnyMessage, AIMessage, HumanMessage
+from typing import Any
+
+from langchain_core.messages import AIMessage, AnyMessage, HumanMessage
 
 
-def get_research_topic(messages: List[AnyMessage]) -> str:
-    """
-    Get the research topic from the messages.
-    """
-    # check if request has a history and combine the messages into a single string
+def get_research_topic(messages: list[AnyMessage]) -> str:
+    """Get the research topic from the messages."""
     if len(messages) == 1:
-        research_topic = messages[-1].content
-    else:
-        research_topic = ""
-        for message in messages:
-            if isinstance(message, HumanMessage):
-                research_topic += f"User: {message.content}\n"
-            elif isinstance(message, AIMessage):
-                research_topic += f"Assistant: {message.content}\n"
+        return str(messages[-1].content)
+
+    research_topic = ""
+    for message in messages:
+        if isinstance(message, HumanMessage):
+            research_topic += f"User: {message.content}\n"
+        elif isinstance(message, AIMessage):
+            research_topic += f"Assistant: {message.content}\n"
     return research_topic
 
 
-def resolve_urls(urls_to_resolve: List[Any], id: int) -> Dict[str, str]:
-    """
-    Create a map of the vertex ai search urls (very long) to a short url with a unique id for each url.
-    Ensures each original URL gets a consistent shortened form while maintaining uniqueness.
-    """
-    prefix = f"https://vertexaisearch.cloud.google.com/id/"
-    urls = [site.web.uri for site in urls_to_resolve]
+def deduplicate_and_format_sources(
+    results: list[dict[str, Any]], search_id: int
+) -> tuple[list[dict[str, str]], str]:
+    """Convert Tavily results into source metadata and a markdown research summary."""
+    sources: list[dict[str, str]] = []
+    formatted_sections: list[str] = []
+    seen_urls: set[str] = set()
 
-    # Create a dictionary that maps each unique URL to its first occurrence index
-    resolved_map = {}
-    for idx, url in enumerate(urls):
-        if url not in resolved_map:
-            resolved_map[url] = f"{prefix}{id}-{idx}"
+    for idx, result in enumerate(results):
+        url = result.get("url")
+        if not url or url in seen_urls:
+            continue
 
-    return resolved_map
+        seen_urls.add(url)
+        title = result.get("title") or f"Source {idx + 1}"
+        short_url = f"https://search.local/{search_id}-{idx}"
+        content = (result.get("content") or "").strip()
 
-
-def insert_citation_markers(text, citations_list):
-    """
-    Inserts citation markers into a text string based on start and end indices.
-
-    Args:
-        text (str): The original text string.
-        citations_list (list): A list of dictionaries, where each dictionary
-                               contains 'start_index', 'end_index', and
-                               'segment_string' (the marker to insert).
-                               Indices are assumed to be for the original text.
-
-    Returns:
-        str: The text with citation markers inserted.
-    """
-    # Sort citations by end_index in descending order.
-    # If end_index is the same, secondary sort by start_index descending.
-    # This ensures that insertions at the end of the string don't affect
-    # the indices of earlier parts of the string that still need to be processed.
-    sorted_citations = sorted(
-        citations_list, key=lambda c: (c["end_index"], c["start_index"]), reverse=True
-    )
-
-    modified_text = text
-    for citation_info in sorted_citations:
-        # These indices refer to positions in the *original* text,
-        # but since we iterate from the end, they remain valid for insertion
-        # relative to the parts of the string already processed.
-        end_idx = citation_info["end_index"]
-        marker_to_insert = ""
-        for segment in citation_info["segments"]:
-            marker_to_insert += f" [{segment['label']}]({segment['short_url']})"
-        # Insert the citation marker at the original end_idx position
-        modified_text = (
-            modified_text[:end_idx] + marker_to_insert + modified_text[end_idx:]
+        sources.append(
+            {
+                "label": title,
+                "short_url": short_url,
+                "value": url,
+            }
         )
 
-    return modified_text
-
-
-def get_citations(response, resolved_urls_map):
-    """
-    Extracts and formats citation information from a Gemini model's response.
-
-    This function processes the grounding metadata provided in the response to
-    construct a list of citation objects. Each citation object includes the
-    start and end indices of the text segment it refers to, and a string
-    containing formatted markdown links to the supporting web chunks.
-
-    Args:
-        response: The response object from the Gemini model, expected to have
-                  a structure including `candidates[0].grounding_metadata`.
-                  It also relies on a `resolved_map` being available in its
-                  scope to map chunk URIs to resolved URLs.
-
-    Returns:
-        list: A list of dictionaries, where each dictionary represents a citation
-              and has the following keys:
-              - "start_index" (int): The starting character index of the cited
-                                     segment in the original text. Defaults to 0
-                                     if not specified.
-              - "end_index" (int): The character index immediately after the
-                                   end of the cited segment (exclusive).
-              - "segments" (list[str]): A list of individual markdown-formatted
-                                        links for each grounding chunk.
-              - "segment_string" (str): A concatenated string of all markdown-
-                                        formatted links for the citation.
-              Returns an empty list if no valid candidates or grounding supports
-              are found, or if essential data is missing.
-    """
-    citations = []
-
-    # Ensure response and necessary nested structures are present
-    if not response or not response.candidates:
-        return citations
-
-    candidate = response.candidates[0]
-    if (
-        not hasattr(candidate, "grounding_metadata")
-        or not candidate.grounding_metadata
-        or not hasattr(candidate.grounding_metadata, "grounding_supports")
-    ):
-        return citations
-
-    for support in candidate.grounding_metadata.grounding_supports:
-        citation = {}
-
-        # Ensure segment information is present
-        if not hasattr(support, "segment") or support.segment is None:
-            continue  # Skip this support if segment info is missing
-
-        start_index = (
-            support.segment.start_index
-            if support.segment.start_index is not None
-            else 0
+        formatted_sections.append(
+            "\n".join(
+                [
+                    f"- {title} [{title}]({short_url})",
+                    f"  URL: {url}",
+                    f"  Snippet: {content or 'No snippet returned.'}",
+                ]
+            )
         )
 
-        # Ensure end_index is present to form a valid segment
-        if support.segment.end_index is None:
-            continue  # Skip if end_index is missing, as it's crucial
+    if not formatted_sections:
+        formatted_sections.append("- No Tavily results were returned for this query.")
 
-        # Add 1 to end_index to make it an exclusive end for slicing/range purposes
-        # (assuming the API provides an inclusive end_index)
-        citation["start_index"] = start_index
-        citation["end_index"] = support.segment.end_index
-
-        citation["segments"] = []
-        if (
-            hasattr(support, "grounding_chunk_indices")
-            and support.grounding_chunk_indices
-        ):
-            for ind in support.grounding_chunk_indices:
-                try:
-                    chunk = candidate.grounding_metadata.grounding_chunks[ind]
-                    resolved_url = resolved_urls_map.get(chunk.web.uri, None)
-                    citation["segments"].append(
-                        {
-                            "label": chunk.web.title.split(".")[:-1][0],
-                            "short_url": resolved_url,
-                            "value": chunk.web.uri,
-                        }
-                    )
-                except (IndexError, AttributeError, NameError):
-                    # Handle cases where chunk, web, uri, or resolved_map might be problematic
-                    # For simplicity, we'll just skip adding this particular segment link
-                    # In a production system, you might want to log this.
-                    pass
-        citations.append(citation)
-    return citations
+    return sources, "\n".join(formatted_sections)

docker-compose.yml

@@ -28,7 +28,7 @@ services:
       retries: 5
       interval: 5s
   langgraph-api:
-    image: gemini-fullstack-langgraph
+    image: openai-fullstack-langgraph
     container_name: langgraph-api
     ports:
       - "8123:8000"
@@ -38,7 +38,9 @@ services:
       langgraph-postgres:
         condition: service_healthy
     environment:
-      GEMINI_API_KEY: ${GEMINI_API_KEY}
+      OPENAI_API_KEY: ${OPENAI_API_KEY}
+      OPENAI_BASE_URL: ${OPENAI_BASE_URL:-https://open.bigmodel.cn/api/paas/v4}
+      TAVILY_API_KEY: ${TAVILY_API_KEY}
       LANGSMITH_API_KEY: ${LANGSMITH_API_KEY}
       REDIS_URI: redis://langgraph-redis:6379
       POSTGRES_URI: postgres://postgres:postgres@langgraph-postgres:5432/postgres?sslmode=disable

frontend/package-lock.json

@@ -2259,6 +2259,7 @@
       "integrity": "sha512-wIX2aSZL5FE+MR0JlvF87BNVrtFWf6AE6rxSE9X7OwnVvoyCQjpzSRJ+M87se/4QCkCiebQAqrJ0y6fwIyi7nw==",
       "devOptional": true,
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "undici-types": "~6.21.0"
       }
@@ -2268,6 +2269,7 @@
       "resolved": "https://registry.npmjs.org/@types/react/-/react-19.1.2.tgz",
       "integrity": "sha512-oxLPMytKchWGbnQM9O7D67uPa9paTNxO7jVoNMXgkkErULBPhPARCfkKL9ytcIJJRGjbsVwW4ugJzyFFvm/Tiw==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "csstype": "^3.0.2"
       }
@@ -2278,6 +2280,7 @@
       "integrity": "sha512-rJXC08OG0h3W6wDMFxQrZF00Kq6qQvw0djHRdzl3U5DnIERz0MRce3WVc7IS6JYBwtaP/DwYtRRjVlvivNveKg==",
       "devOptional": true,
       "license": "MIT",
+      "peer": true,
       "peerDependencies": {
         "@types/react": "^19.0.0"
       }
@@ -2336,6 +2339,7 @@
       "integrity": "sha512-oU/OtYVydhXnumd0BobL9rkJg7wFJ9bFFPmSmB/bf/XWN85hlViji59ko6bSKBXyseT9V8l+CN1nwmlbiN0G7Q==",
       "dev": true,
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "@typescript-eslint/scope-manager": "8.31.1",
         "@typescript-eslint/types": "8.31.1",
@@ -2531,6 +2535,7 @@
       "integrity": "sha512-OvQ/2pUDKmgfCg++xsTX1wGxfTaszcHVcTctW4UJB4hibJx2HXxxO5UmVgyjMa+ZDsiaf5wWLXYpRWMmBI0QHg==",
       "dev": true,
       "license": "MIT",
+      "peer": true,
       "bin": {
         "acorn": "bin/acorn"
       },
@@ -2998,6 +3003,7 @@
       "integrity": "sha512-E6Mtz9oGQWDCpV12319d59n4tx9zOTXSTmc8BLVxBx+G/0RdM5MvEEJLU9c0+aleoePYYgVTOsRblx433qmhWQ==",
       "dev": true,
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "@eslint-community/eslint-utils": "^4.2.0",
         "@eslint-community/regexpp": "^4.12.1",
@@ -4879,6 +4885,7 @@
       "resolved": "https://registry.npmjs.org/react/-/react-19.1.0.tgz",
       "integrity": "sha512-FS+XFBNvn3GTAWq26joslQgWNoFu08F4kl0J4CgdNKADkdSGXQyTCnKteIAJy96Br6YbpEU1LSzV5dYtjMkMDg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=0.10.0"
       }
@@ -4888,6 +4895,7 @@
       "resolved": "https://registry.npmjs.org/react-dom/-/react-dom-19.1.0.tgz",
       "integrity": "sha512-Xs1hdnE+DyKgeHJeJznQmYMIBG3TKIHJJT95Q58nHLSrElKlGQqDTR2HQ9fx5CN/Gk6Vh/kupBTDLU11/nDk/g==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "scheduler": "^0.26.0"
       },
@@ -5345,6 +5353,7 @@
       "resolved": "https://registry.npmjs.org/picomatch/-/picomatch-4.0.2.tgz",
       "integrity": "sha512-M7BAV6Rlcy5u+m6oPhAPFgJTzAioX/6B0DxyvDlo9l8+T3nLKbrczg2WLUyzd45L8RqfUMyGPzekbMvX2Ldkwg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=12"
       },
@@ -5439,6 +5448,7 @@
       "integrity": "sha512-84MVSjMEHP+FQRPy3pX9sTVV/INIex71s9TL2Gm5FG/WG1SqXeKyZ0k7/blY/4FdOzI12CBy1vGc4og/eus0fw==",
       "dev": true,
       "license": "Apache-2.0",
+      "peer": true,
       "bin": {
         "tsc": "bin/tsc",
         "tsserver": "bin/tsserver"
@@ -5663,6 +5673,7 @@
       "resolved": "https://registry.npmjs.org/vite/-/vite-6.3.4.tgz",
       "integrity": "sha512-BiReIiMS2fyFqbqNT/Qqt4CVITDU9M9vE+DKcVAsB+ZV0wvTKd+3hMbkpxz1b+NmEDMegpVbisKiAZOnvO92Sw==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "esbuild": "^0.25.0",
         "fdir": "^6.4.4",
@@ -5751,6 +5762,7 @@
       "resolved": "https://registry.npmjs.org/picomatch/-/picomatch-4.0.2.tgz",
       "integrity": "sha512-M7BAV6Rlcy5u+m6oPhAPFgJTzAioX/6B0DxyvDlo9l8+T3nLKbrczg2WLUyzd45L8RqfUMyGPzekbMvX2Ldkwg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=12"
       },
@@ -5802,6 +5814,7 @@
       "resolved": "https://registry.npmjs.org/zod/-/zod-3.24.4.tgz",
       "integrity": "sha512-OdqJE9UDRPwWsrHjLN2F8bPxvwJBK22EHLWtanu0LSYr5YqzsaaW3RMgmjwr8Rypg5k+meEJdSPXJZXE/yqOMg==",
       "license": "MIT",
+      "peer": true,
       "funding": {
         "url": "https://github.com/sponsors/colinhacks"
       }

frontend/src/components/InputForm.tsx

@@ -26,7 +26,7 @@ export const InputForm: React.FC<InputFormProps> = ({
 }) => {
   const [internalInputValue, setInternalInputValue] = useState("");
   const [effort, setEffort] = useState("medium");
-  const [model, setModel] = useState("gemini-2.5-flash-preview-04-17");
+  const [model, setModel] = useState("GLM-4.5-Air");
 
   const handleInternalSubmit = (e?: React.FormEvent) => {
     if (e) e.preventDefault();
@@ -35,10 +35,9 @@ export const InputForm: React.FC<InputFormProps> = ({
     setInternalInputValue("");
   };
 
-  const handleInternalKeyDown = (
-    e: React.KeyboardEvent<HTMLTextAreaElement>
-  ) => {
-    if (e.key === "Enter" && !e.shiftKey) {
+  const handleKeyDown = (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
+    // Submit with Ctrl+Enter (Windows/Linux) or Cmd+Enter (Mac)
+    if (e.key === "Enter" && (e.ctrlKey || e.metaKey)) {
       e.preventDefault();
       handleInternalSubmit();
     }
@@ -59,9 +58,9 @@ export const InputForm: React.FC<InputFormProps> = ({
         <Textarea
           value={internalInputValue}
           onChange={(e) => setInternalInputValue(e.target.value)}
-          onKeyDown={handleInternalKeyDown}
+          onKeyDown={handleKeyDown}
           placeholder="Who won the Euro 2024 and scored the most goals?"
-          className={`w-full text-neutral-100 placeholder-neutral-500 resize-none border-0 focus:outline-none focus:ring-0 outline-none focus-visible:ring-0 shadow-none 
+          className={`w-full text-neutral-100 placeholder-neutral-500 resize-none border-0 focus:outline-none focus:ring-0 outline-none focus-visible:ring-0 shadow-none
                         md:text-base  min-h-[56px] max-h-[200px]`}
           rows={1}
         />
@@ -137,27 +136,11 @@ export const InputForm: React.FC<InputFormProps> = ({
               </SelectTrigger>
               <SelectContent className="bg-neutral-700 border-neutral-600 text-neutral-300 cursor-pointer">
                 <SelectItem
-                  value="gemini-2.0-flash"
+                  value="GLM-4.5-Air"
                   className="hover:bg-neutral-600 focus:bg-neutral-600 cursor-pointer"
                 >
                   <div className="flex items-center">
-                    <Zap className="h-4 w-4 mr-2 text-yellow-400" /> 2.0 Flash
-                  </div>
-                </SelectItem>
-                <SelectItem
-                  value="gemini-2.5-flash-preview-04-17"
-                  className="hover:bg-neutral-600 focus:bg-neutral-600 cursor-pointer"
-                >
-                  <div className="flex items-center">
-                    <Zap className="h-4 w-4 mr-2 text-orange-400" /> 2.5 Flash
-                  </div>
-                </SelectItem>
-                <SelectItem
-                  value="gemini-2.5-pro-preview-05-06"
-                  className="hover:bg-neutral-600 focus:bg-neutral-600 cursor-pointer"
-                >
-                  <div className="flex items-center">
-                    <Cpu className="h-4 w-4 mr-2 text-purple-400" /> 2.5 Pro
+                    <Zap className="h-4 w-4 mr-2 text-cyan-400" /> GLM-4.5-Air
                   </div>
                 </SelectItem>
               </SelectContent>

frontend/src/components/WelcomeScreen.tsx

@@ -33,7 +33,7 @@ export const WelcomeScreen: React.FC<WelcomeScreenProps> = ({
       />
     </div>
     <p className="text-xs text-neutral-500">
-      Powered by Google Gemini and LangChain LangGraph.
+      Powered by GLM-4.5-Air, Tavily, and LangChain LangGraph.
     </p>
   </div>
 );

frontend/vite.config.ts

@@ -9,7 +9,7 @@ export default defineConfig({
   base: "/app/",
   resolve: {
     alias: {
-      "@": path.resolve(new URL(".", import.meta.url).pathname, "./src"),
+      "@": path.resolve(__dirname, "./src"),
     },
   },
   server: {