diff --git a/backend/context.py b/backend/context.py
new file mode 100644
index 000000000..e413cb0a8
--- /dev/null
+++ b/backend/context.py
@@ -0,0 +1,89 @@
+"""Context management for multi-message conversations with smart summarization."""
+
+from typing import List, Dict, Any
+from .openrouter import query_model
+
+
+async def summarize_older_messages(messages: List[Dict[str, str]]) -> str:
+ """Summarize older messages into a concise conversation summary."""
+ conversation_text = ""
+ for msg in messages:
+ role = msg["role"].capitalize()
+ if msg["role"] == "user":
+ conversation_text += f"{role}: {msg['content']}\n\n"
+ else:
+ if "stage3" in msg and "response" in msg["stage3"]:
+ conversation_text += f"{role}: {msg['stage3']['response']}\n\n"
+ elif "content" in msg:
+ conversation_text += f"{role}: {msg['content']}\n\n"
+
+ summary_prompt = f"""Summarize the following conversation concisely in 2-3 sentences. Focus on key topics, questions asked, and important context that would be needed to understand follow-up questions.
+
+Conversation:
+{conversation_text}
+
+Concise summary:"""
+
+ messages_for_api = [{"role": "user", "content": summary_prompt}]
+ response = await query_model("google/gemini-2.0-flash-exp:free", messages_for_api, timeout=30.0)
+
+ if response is None:
+ return "Previous conversation: " + conversation_text[:200] + "..."
+
+ return response.get('content', '').strip()
+
+
+def format_assistant_message(assistant_msg: Dict[str, Any]) -> str:
+ """Convert council's 3-stage output into clean text for context."""
+ if "stage3" in assistant_msg and "response" in assistant_msg["stage3"]:
+ return assistant_msg["stage3"]["response"]
+
+ if "content" in assistant_msg:
+ return assistant_msg["content"]
+
+ return "[Assistant response]"
+
+
+async def build_context_messages(
+ conversation_messages: List[Dict[str, Any]],
+ current_query: str,
+ recent_message_limit: int = 5
+) -> List[Dict[str, str]]:
+ """Build message history with smart summarization for long conversations."""
+ if len(conversation_messages) == 0:
+ return [{"role": "user", "content": current_query}]
+
+ num_recent_messages = recent_message_limit * 2
+
+ if len(conversation_messages) <= num_recent_messages:
+ formatted_messages = []
+ for msg in conversation_messages:
+ if msg["role"] == "user":
+ formatted_messages.append({"role": "user", "content": msg["content"]})
+ else:
+ content = format_assistant_message(msg)
+ formatted_messages.append({"role": "assistant", "content": content})
+
+ formatted_messages.append({"role": "user", "content": current_query})
+ return formatted_messages
+
+ older_messages = conversation_messages[:-num_recent_messages]
+ recent_messages = conversation_messages[-num_recent_messages:]
+
+ summary = await summarize_older_messages(older_messages)
+
+ formatted_messages = [
+ {"role": "user", "content": f"[Previous conversation summary: {summary}]"},
+ {"role": "assistant", "content": "I understand the previous conversation context."}
+ ]
+
+ for msg in recent_messages:
+ if msg["role"] == "user":
+ formatted_messages.append({"role": "user", "content": msg["content"]})
+ else:
+ content = format_assistant_message(msg)
+ formatted_messages.append({"role": "assistant", "content": content})
+
+ formatted_messages.append({"role": "user", "content": current_query})
+ return formatted_messages
+
diff --git a/backend/council.py b/backend/council.py
index 5069abec9..a2579fbde 100644
--- a/backend/council.py
+++ b/backend/council.py
@@ -5,19 +5,16 @@
from .config import COUNCIL_MODELS, CHAIRMAN_MODEL
-async def stage1_collect_responses(user_query: str) -> List[Dict[str, Any]]:
+async def stage1_collect_responses(messages: List[Dict[str, str]]) -> List[Dict[str, Any]]:
"""
Stage 1: Collect individual responses from all council models.
Args:
- user_query: The user's question
+ messages: Full message history including current query
Returns:
List of dicts with 'model' and 'response' keys
"""
- messages = [{"role": "user", "content": user_query}]
-
- # Query all models in parallel
responses = await query_models_parallel(COUNCIL_MODELS, messages)
# Format results
@@ -293,35 +290,34 @@ async def generate_conversation_title(user_query: str) -> str:
return title
-async def run_full_council(user_query: str) -> Tuple[List, List, Dict, Dict]:
+async def run_full_council(messages: List[Dict[str, str]]) -> Tuple[List, List, Dict, Dict]:
"""
- Run the complete 3-stage council process.
+ Run the complete 3-stage council process with conversation context.
Args:
- user_query: The user's question
+ messages: Full message history in OpenAI format
Returns:
Tuple of (stage1_results, stage2_results, stage3_result, metadata)
"""
- # Stage 1: Collect individual responses
- stage1_results = await stage1_collect_responses(user_query)
+ current_query = messages[-1]["content"]
+
+ stage1_results = await stage1_collect_responses(messages)
- # If no models responded successfully, return error
if not stage1_results:
return [], [], {
"model": "error",
"response": "All models failed to respond. Please try again."
}, {}
- # Stage 2: Collect rankings
- stage2_results, label_to_model = await stage2_collect_rankings(user_query, stage1_results)
+ stage2_results, label_to_model = await stage2_collect_rankings(current_query, stage1_results)
# Calculate aggregate rankings
aggregate_rankings = calculate_aggregate_rankings(stage2_results, label_to_model)
# Stage 3: Synthesize final answer
stage3_result = await stage3_synthesize_final(
- user_query,
+ current_query,
stage1_results,
stage2_results
)
diff --git a/backend/main.py b/backend/main.py
index e33ce59a6..f36fa7ae9 100644
--- a/backend/main.py
+++ b/backend/main.py
@@ -11,6 +11,7 @@
from . import storage
from .council import run_full_council, generate_conversation_title, stage1_collect_responses, stage2_collect_rankings, stage3_synthesize_final, calculate_aggregate_rankings
+from .context import build_context_messages
app = FastAPI(title="LLM Council API")
@@ -101,11 +102,13 @@ async def send_message(conversation_id: str, request: SendMessageRequest):
title = await generate_conversation_title(request.content)
storage.update_conversation_title(conversation_id, title)
- # Run the 3-stage council process
- stage1_results, stage2_results, stage3_result, metadata = await run_full_council(
+ messages = await build_context_messages(
+ conversation["messages"][:-1],
request.content
)
+ stage1_results, stage2_results, stage3_result, metadata = await run_full_council(messages)
+
# Add assistant message with all stages
storage.add_assistant_message(
conversation_id,
@@ -147,9 +150,14 @@ async def event_generator():
if is_first_message:
title_task = asyncio.create_task(generate_conversation_title(request.content))
+ messages = await build_context_messages(
+ conversation["messages"][:-1],
+ request.content
+ )
+
# Stage 1: Collect responses
yield f"data: {json.dumps({'type': 'stage1_start'})}\n\n"
- stage1_results = await stage1_collect_responses(request.content)
+ stage1_results = await stage1_collect_responses(messages)
yield f"data: {json.dumps({'type': 'stage1_complete', 'data': stage1_results})}\n\n"
# Stage 2: Collect rankings
diff --git a/frontend/.nvmrc b/frontend/.nvmrc
new file mode 100644
index 000000000..db2b6f4ba
--- /dev/null
+++ b/frontend/.nvmrc
@@ -0,0 +1,9 @@
+20
+
+
+
+
+
+
+
+
diff --git a/frontend/src/components/ChatInterface.css b/frontend/src/components/ChatInterface.css
index 0d013003a..715a825fb 100644
--- a/frontend/src/components/ChatInterface.css
+++ b/frontend/src/components/ChatInterface.css
@@ -33,6 +33,21 @@
font-size: 16px;
}
+.context-indicator {
+ display: flex;
+ align-items: center;
+ gap: 8px;
+ padding: 10px 16px;
+ margin-bottom: 20px;
+ background: linear-gradient(135deg, #f8f9ff 0%, #f0f7ff 100%);
+ border: 1px solid #d0e7ff;
+ border-radius: 8px;
+ color: #4a90e2;
+ font-size: 13px;
+ font-weight: 500;
+ box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05);
+}
+
.message-group {
margin-bottom: 32px;
}
diff --git a/frontend/src/components/ChatInterface.jsx b/frontend/src/components/ChatInterface.jsx
index 3ae796caa..544cdb98d 100644
--- a/frontend/src/components/ChatInterface.jsx
+++ b/frontend/src/components/ChatInterface.jsx
@@ -57,7 +57,13 @@ export default function ChatInterface({
Ask a question to consult the LLM Council
) : (
- conversation.messages.map((msg, index) => (
+ <>
+ {conversation.messages.length > 6 && (
+
+ 💠Using conversation context ({conversation.messages.length} messages)
+
+ )}
+ {conversation.messages.map((msg, index) => (
{msg.role === 'user' ? (
@@ -107,7 +113,8 @@ export default function ChatInterface({
)}