milestone v1: works so far

config.yaml

@@ -13,7 +13,7 @@ openrouter_base_url: "https://openrouter.ai/api/v1"
 llm_model: "google/gemini-2.0-flash-001"
 
 # Number of hypotheses to generate
-num_hypotheses: 3
+num_hypotheses: 6
 
 # Elo K-factor
 elo_k_factor: 32
@@ -32,3 +32,8 @@ fastapi_host: "0.0.0.0"
 
 # FastAPI port
 fastapi_port: 8000
+
+# Temperature settings for each step
+step_temperatures:
+  generation: 0.7
+  reflection: 0.5

graph_visualizer.py

@@ -0,0 +1,35 @@
+import networkx as nx
+import matplotlib.pyplot as plt
+
+def visualize_graph(adjacency_graph):
+    """Visualizes an adjacency graph using networkx and matplotlib."""
+
+    graph = nx.DiGraph()  # Use DiGraph for directed graph
+
+    for node, edges in adjacency_graph.items():
+        for edge in edges:
+            graph.add_edge(node, edge['other_id'], weight=edge['similarity'])
+
+    pos = nx.spring_layout(graph)  # Node positioning for visualization
+
+    # Get edge weights and normalize them
+    weights = [edge['weight'] for u, v, edge in graph.edges(data=True)]
+    normalized_weights = [(w - min(weights)) / (max(weights) - min(weights)) for w in weights]
+
+    # Create a color map
+    cmap = plt.cm.viridis
+
+    # Map normalized weights to colors
+    edge_colors = [cmap(w) for w in normalized_weights]
+    
+    nx.draw(graph, pos, with_labels=True, node_size=1500, node_color="skyblue", font_size=10, edge_color=edge_colors)
+    edge_labels = nx.get_edge_attributes(graph, 'weight')
+    nx.draw_networkx_edge_labels(graph, pos, edge_labels=edge_labels, font_size=8)
+
+
+    plt.title("Adjacency Graph")
+    plt.show()
+
+if __name__ == "__main__":
+    adjacency_graph = {"G3250":[{"other_id":"E2029","similarity":0.8886164931432184},{"other_id":"G4687","similarity":0.7799796722164661},{"other_id":"G2491","similarity":0.7358682896118928}],"E2029":[{"other_id":"G3250","similarity":0.589185898055919},{"other_id":"G4687","similarity":0.5547903202019775},{"other_id":"G2491","similarity":0.4763465778429552}],"G4687":[{"other_id":"G3250","similarity":0.8534847661087587},{"other_id":"E2029","similarity":0.382888810662511},{"other_id":"G2491","similarity":0.9591597530883424}],"G2491":[{"other_id":"G3250","similarity":0.11935305775711214},{"other_id":"E2029","similarity":0.3629634156202275},{"other_id":"G4687","similarity":0.810511185411589}]}
+    visualize_graph(adjacency_graph)

proposal-gen-v1.py

@@ -38,6 +38,9 @@ def load_config(config_path: str) -> Dict:
     except AttributeError as e:
         print("Error: Invalid logging level in config file")
         exit(1)
+    except KeyError as e:
+        print(f"Error: Missing key in config file: {e}")
+        exit(1)
 
 
 def setup_logger(log_filename):
@@ -57,10 +60,17 @@ def setup_logger(log_filename):
 # Load configuration at the start
 config = load_config("config.yaml")
 
-def call_llm(prompt: str) -> str:
+def call_llm(prompt: str, temperature: float = 0.7) -> str:
     """
     Calls an LLM via the OpenRouter API and returns the response.
 
+    Args:
+        prompt (str): The input prompt for the LLM.
+        temperature (float, optional): The temperature setting for the LLM. Defaults to 0.7.
+
+    Returns:
+        str: The LLM's response.
+
     Args:
         prompt (str): The input prompt for the LLM.
 
@@ -76,6 +86,7 @@ def call_llm(prompt: str) -> str:
         completion = client.chat.completions.create(
             model=config["llm_model"],
             messages=[{"role": "user", "content": prompt}],
+            temperature=temperature,  # Pass temperature to the API call
         )
     except Exception as e:
         retries = config.get("max_retries", 3)
@@ -96,6 +107,7 @@ def call_llm(prompt: str) -> str:
                 completion = client.chat.completions.create(
                     model=config["llm_model"],
                     messages=[{"role": "user", "content": prompt}],
+                    temperature=temperature,  # Pass temperature to the API call
                 )
                 if completion.choices and len(completion.choices) > 0:
                     return completion.choices[0].message.content
@@ -291,7 +303,8 @@ def call_llm_for_generation(prompt: str, num_hypotheses: int = 3) -> List[Dict]:
     # Modify the prompt to request JSON output
     prompt += "\n\nPlease return the response as a JSON array of objects, where each object has a 'title' and 'text' key."
 
-    response = call_llm(prompt)
+    # Call LLM with the appropriate temperature
+    response = call_llm(prompt, temperature=config["step_temperatures"]["generation"])
     logger.info("LLM response: %s", response)
 
     if "API call failed" in response:
@@ -341,7 +354,8 @@ def call_llm_for_reflection(hypothesis_text: str) -> Dict:
         f"Return the response as a JSON object with the following keys: 'novelty_review', 'feasibility_review', 'comment', 'references'."
 
     )
-    response = call_llm(prompt)
+    # Call LLM with the appropriate temperature
+    response = call_llm(prompt, temperature=config["step_temperatures"]["reflection"])
     logger.info("LLM reflection for hypothesis: %s, response: %s", hypothesis_text, response)
 
     if "API call failed" in response:
@@ -422,7 +436,19 @@ def run_pairwise_debate(hypoA: Hypothesis, hypoB: Hypothesis) -> Hypothesis:
                  the novelty and feasibility scores.
         """
         mapping = {"HIGH": 3, "MEDIUM": 2, "LOW": 1, None: 0}
-        return mapping.get(h.novelty_review, 0) + mapping.get(h.feasibility_review, 0)
+        score_novelty = 0
+        if isinstance(h.novelty_review, str):
+            score_novelty = mapping.get(h.novelty_review, 0)
+        else:
+            logger.error(f"Invalid novelty_review type: {type(h.novelty_review)}, value: {h.novelty_review}")
+
+        score_feasibility = 0
+        if isinstance(h.feasibility_review, str):
+            score_feasibility = mapping.get(h.feasibility_review, 0)
+        else:
+            logger.error(f"Invalid feasibility_review type: {type(h.feasibility_review)}, value: {h.feasibility_review}")
+
+        return score_novelty + score_feasibility
     scoreA = score(hypoA)
     scoreB = score(hypoB)
     winner = hypoA if scoreA > scoreB else hypoB if scoreB > scoreA else random.choice([hypoA, hypoB])