add a config.yaml file

README.md

@@ -46,7 +46,7 @@ The `proposal-gen-v1.py` script implements a multi-agent system that iteratively
     uvicorn proposal-gen-v1:app --host 0.0.0.0 --port 8000
     ```
 4.  **Access the Web Interface:**
-    Open a web browser and go to `http://localhost:8000`.
+    Open a web browser and go to `http://localhost:8000`. (Note: The server log may show `http://0.0.0.0:8000`, which means the server is listening on all network interfaces. However, you should use `localhost` in your browser to access the server from your local machine. You cannot directly type `0.0.0.0` into your browser's address bar.)
 5.  **Enter Research Goal:**
     Enter your research goal in the text area provided.
 6.  **Submit and Run:**
@@ -66,6 +66,28 @@ The system will generate a list of hypotheses related to the research goal. Each
 
 The web interface will display the top-ranked hypotheses after each cycle, along with a meta-review critique and suggested next steps. The results are iterative, meaning that the hypotheses should improve over multiple cycles. Log files are created in the `results/` directory for each run.
 
+## Configuration (config.yaml)
+
+The `config.yaml` file contains settings that control the behavior of the AI Co-Scientist system. Here's a detailed explanation of each option:
+
+*   **`openrouter_base_url`**:  This specifies the base URL for the OpenRouter API.  OpenRouter acts as a proxy to various LLMs, providing a consistent interface.  The default value is `"https://openrouter.ai/api/v1"`, which should work without modification.
+
+*   **`llm_model`**: This setting determines which Large Language Model (LLM) the system will use.  The default is `"google/gemini-2.0-flash-thinking-exp:free"`, which is a free model from Google, hosted on OpenRouter. You can change this to use a different model, but ensure it's compatible with the OpenRouter API and the system's prompts.  Refer to the OpenRouter documentation for available models and their identifiers.
+
+*   **`num_hypotheses`**: This controls the number of initial hypotheses generated in each cycle. The default value is `3`. Increasing this number will explore a broader range of ideas, but may also increase processing time and API costs (if using a paid LLM).
+
+*   **`elo_k_factor`**: This parameter, used in the Elo rating system, determines how much the Elo scores change after each comparison between hypotheses.  A higher `elo_k_factor` (default is `32`) means that scores will change more dramatically, making the ranking more sensitive to individual comparisons.  A lower value will result in slower, more gradual changes in ranking.
+
+*   **`top_k_hypotheses`**: This setting specifies how many of the top-ranked hypotheses are used by the `EvolutionAgent` to create new hypotheses. The default is `2`.  Increasing this value might lead to more diverse combinations, but could also dilute the influence of the very best hypotheses.
+
+*   **`logging_level`**:  This controls the verbosity of the logging output.  Valid values are `"DEBUG"`, `"INFO"`, `"WARNING"`, `"ERROR"`, and `"CRITICAL"`.  The default is `"INFO"`.  `"DEBUG"` provides the most detailed information, while `"CRITICAL"` only logs the most severe errors.
+
+*   **`log_file_name`**: This is the base name for the log files (without the extension). Log files are stored in the `results/` directory. The default is `"app"`. The system automatically adds a timestamp and the `.txt` extension to the log file name (e.g., `app_2025-02-28_09-18-00.txt`).
+
+*   **`fastapi_host`**: This setting controls the network interface that the FastAPI application will listen on. The default value, `"0.0.0.0"`, makes the application accessible from any network interface, including your local machine and potentially other computers on your network.  You could change this to `"127.0.0.1"` to restrict access to only your local machine.
+
+*   **`fastapi_port`**: This specifies the port number that the FastAPI application will use. The default is `8000`. You can change this if you have another application already using port 8000 or if you prefer a different port for other reasons.
+
 ## Known Limitations
 
 *   **LLM Dependency:** The quality of the results heavily depends on the capabilities of the underlying LLM.

proposal-gen-v1.py

@@ -8,8 +8,10 @@ from openai import OpenAI
 import os
 import datetime
 from fastapi import FastAPI, HTTPException, responses
+from fastapi.staticfiles import StaticFiles
 from pydantic import BaseModel
 import uvicorn
+import yaml
 
 # Configure logging for production readiness.
 # logging.basicConfig(
@@ -19,20 +21,42 @@ import uvicorn
 # )
 # logger = logging.getLogger("co_scientist") # global logger
 
+def load_config(config_path: str) -> Dict:
+    """Loads the configuration from the specified YAML file."""
+    try:
+        with open(config_path, "r") as f:
+            config = yaml.safe_load(f)
+            # Convert logging level string to actual level
+            config["logging_level"] = getattr(logging, config["logging_level"].upper(), logging.INFO)
+        return config
+    except FileNotFoundError:
+        print(f"Error: Configuration file not found at {config_path}")
+        exit(1)
+    except yaml.YAMLError as e:
+        print(f"Error parsing YAML in {config_path}: {e}")
+        exit(1)
+    except AttributeError as e:
+        print("Error: Invalid logging level in config file")
+        exit(1)
+
+
 def setup_logger(log_filename):
     logger = logging.getLogger(log_filename)  # Create a logger with the filename
-    logger.setLevel(logging.INFO)
+    logger.setLevel(config["logging_level"])
     formatter = logging.Formatter("%(asctime)s %(levelname)s %(name)s: %(message)s")
 
     # Remove existing handlers to avoid duplicate logs
     for handler in logger.handlers[:]:
         logger.removeHandler(handler)
 
-    file_handler = logging.FileHandler(log_filename)
+    file_handler = logging.FileHandler(f"{config['log_file_name']}_{log_filename}")
     file_handler.setFormatter(formatter)
     logger.addHandler(file_handler)
     return logger
 
+# Load configuration at the start
+config = load_config("config.yaml")
+
 def call_llm(prompt: str) -> str:
     """
     Calls an LLM via the OpenRouter API and returns the response.
@@ -44,16 +68,14 @@ def call_llm(prompt: str) -> str:
         str: The LLM's response.
     """
     client = OpenAI(
-      base_url="https://openrouter.ai/api/v1",
-      api_key=os.getenv("OPENROUTER_API_KEY")
+        base_url=config["openrouter_base_url"],
+        api_key=os.getenv("OPENROUTER_API_KEY"),
     )
 
     try:
         completion = client.chat.completions.create(
-            model="google/gemini-2.0-flash-thinking-exp:free",  # Or any other suitable model
-            messages=[
-                {"role": "user", "content": prompt}
-            ],
+            model=config["llm_model"],
+            messages=[{"role": "user", "content": prompt}],
         )
     except Exception as e:
         # If the library raises an exception (e.g., for invalid key, rate limit, etc.)
@@ -281,7 +303,7 @@ def run_pairwise_debate(hypoA: Hypothesis, hypoB: Hypothesis) -> Hypothesis:
                 hypoA.hypothesis_id, scoreA, hypoB.hypothesis_id, scoreB, winner.hypothesis_id)
     return winner
 
-def update_elo(winner: Hypothesis, loser: Hypothesis, k_factor: int = 32):
+def update_elo(winner: Hypothesis, loser: Hypothesis, k_factor: int = config["elo_k_factor"]):
     """
     Updates the Elo scores of two hypotheses after a pairwise comparison.
 
@@ -355,9 +377,9 @@ class GenerationAgent:
         prompt = (
             f"Research Goal: {research_goal.description}\n"
             f"Constraints: {research_goal.constraints}\n"
-            "Please propose 2 new hypotheses with rationale.\n"
+            f"Please propose {config['num_hypotheses']} new hypotheses with rationale.\n"
         )
-        raw_output = call_llm_for_generation(prompt, num_hypotheses=2)
+        raw_output = call_llm_for_generation(prompt, num_hypotheses=config["num_hypotheses"])
         new_hypos = []
         for idea in raw_output:
             hypo_id = generate_unique_id("G")
@@ -431,7 +453,7 @@ class EvolutionAgent:
         """
         active = context.get_active_hypotheses()
         sorted_by_elo = sorted(active, key=lambda h: h.elo_score, reverse=True)
-        top_candidates = sorted_by_elo[:top_k]
+        top_candidates = sorted_by_elo[:config["top_k_hypotheses"]]
         new_hypotheses = []
         if len(top_candidates) >= 2:
             new_h = combine_hypotheses(top_candidates[0], top_candidates[1])
@@ -567,6 +589,8 @@ global_context = ContextMemory()
 supervisor = SupervisorAgent()
 current_research_goal: Optional[ResearchGoal] = None
 
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
 @app.post("/research_goal", response_model=dict)
 def set_research_goal(goal: ResearchGoalRequest):
     """
@@ -714,4 +738,4 @@ async def root():
 
 if __name__ == "__main__":
     # Run with: uvicorn this_script:app --host 0.0.0.0 --port 8000
-    uvicorn.run("proposal-gen-v1:app", host="0.0.0.0", port=8000, reload=False)
+    uvicorn.run("proposal-gen-v1:app", host=config["fastapi_host"], port=config["fastapi_port"], reload=False)

requirements.txt

@@ -1,4 +1,5 @@
 openai
 fastapi
-pydantic
 uvicorn
+pydantic
+PyYAML