torbjorn/my-pal-mcp-server
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improvements to consensus

Browse Source
This commit is contained in:
Fahad
2025-08-08 12:59:41 +05:00
parent b212cae5de
commit e29deb23db
5 changed files with 32 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
config.py
View File

@@ -14,7 +14,7 @@ import os
# These values are used in server responses and for tracking releases # These values are used in server responses and for tracking releases
# IMPORTANT: This is the single source of truth for version and author info # IMPORTANT: This is the single source of truth for version and author info
# Semantic versioning: MAJOR.MINOR.PATCH # Semantic versioning: MAJOR.MINOR.PATCH
__version__ = "5.8.4" __version__ = "5.8.5"
# Last update date in ISO format # Last update date in ISO format
__updated__ = "2025-08-08" __updated__ = "2025-08-08"
# Primary maintainer # Primary maintainer

6
tests/test_auto_model_planner_fix.py
View File

@@ -100,7 +100,7 @@ class TestAutoModelPlannerFix:
import json import json
response_data = json.loads(result[0].text) response_data = json.loads(result[0].text)
assert response_data["status"] == "planner_complete" assert response_data["status"] == "planning_complete"
assert response_data["step_number"] == 1 assert response_data["step_number"] == 1
@patch("config.DEFAULT_MODEL", "auto") @patch("config.DEFAULT_MODEL", "auto")
@@ -172,7 +172,7 @@ class TestAutoModelPlannerFix:
import json import json
response1 = json.loads(result1[0].text) response1 = json.loads(result1[0].text)
assert response1["status"] == "pause_for_planner" assert response1["status"] == "pause_for_planning"
assert response1["next_step_required"] is True assert response1["next_step_required"] is True
assert "continuation_id" in response1 assert "continuation_id" in response1
@@ -190,7 +190,7 @@ class TestAutoModelPlannerFix:
assert len(result2) > 0 assert len(result2) > 0
response2 = json.loads(result2[0].text) response2 = json.loads(result2[0].text)
assert response2["status"] == "pause_for_planner" assert response2["status"] == "pause_for_planning"
assert response2["step_number"] == 2 assert response2["step_number"] == 2
def test_other_tools_still_require_models(self): def test_other_tools_still_require_models(self):

6
tests/test_planner.py
View File

@@ -226,7 +226,7 @@ class TestPlannerTool:
parsed_response = json.loads(response_text) parsed_response = json.loads(response_text)
# Check final step structure # Check final step structure
assert parsed_response["status"] == "planner_complete" assert parsed_response["status"] == "planning_complete"
assert parsed_response["step_number"] == 10 assert parsed_response["step_number"] == 10
assert parsed_response["planning_complete"] is True assert parsed_response["planning_complete"] is True
assert "plan_summary" in parsed_response assert "plan_summary" in parsed_response
@@ -329,7 +329,7 @@ class TestPlannerTool:
# Total steps should be adjusted to match current step # Total steps should be adjusted to match current step
assert parsed_response["total_steps"] == 8 assert parsed_response["total_steps"] == 8
assert parsed_response["step_number"] == 8 assert parsed_response["step_number"] == 8
assert parsed_response["status"] == "pause_for_planner" assert parsed_response["status"] == "pause_for_planning"
@pytest.mark.asyncio @pytest.mark.asyncio
async def test_execute_error_handling(self): async def test_execute_error_handling(self):
@@ -457,6 +457,6 @@ class TestPlannerToolIntegration:
assert parsed_response["total_steps"] == 3 assert parsed_response["total_steps"] == 3
assert parsed_response["continuation_id"] == "test-simple-uuid" assert parsed_response["continuation_id"] == "test-simple-uuid"
# For simple plans (< 5 steps), expect normal flow without deep thinking pause # For simple plans (< 5 steps), expect normal flow without deep thinking pause
assert parsed_response["status"] == "pause_for_planner" assert parsed_response["status"] == "pause_for_planning"
assert "thinking_required" not in parsed_response assert "thinking_required" not in parsed_response
assert "Continue with step 2" in parsed_response["next_steps"] assert "Continue with step 2" in parsed_response["next_steps"]

34
tools/consensus.py
View File

@@ -37,11 +37,12 @@ logger = logging.getLogger(__name__)
# Tool-specific field descriptions for consensus workflow # Tool-specific field descriptions for consensus workflow
CONSENSUS_WORKFLOW_FIELD_DESCRIPTIONS = { CONSENSUS_WORKFLOW_FIELD_DESCRIPTIONS = {
"step": ( "step": (
"Describe your current consensus analysis step. In step 1, provide your own neutral, balanced analysis " "In step 1: Provide the EXACT question or proposal that ALL models will evaluate. This should be phrased as a clear "
"of the proposal/idea/plan after thinking carefully about all aspects. Consider technical feasibility, " "question or problem statement, NOT as 'I will analyze...' or 'Let me examine...'. For example: 'Should we build a "
"user value, implementation complexity, and alternatives. In subsequent steps (2+), you will receive " "search component in SwiftUI for use in an AppKit app?' or 'Evaluate the proposal to migrate our database from MySQL "
"individual model responses to synthesize. CRITICAL: Be thorough and balanced in your initial assessment, " "to PostgreSQL'. This exact text will be sent to all models for their independent evaluation. "
"considering both benefits and risks, opportunities and challenges." "In subsequent steps (2+): This field is for internal tracking only - you can provide notes about the model response "
"you just received. This will NOT be sent to other models (they all receive the original proposal from step 1)."
), ),
"step_number": ( "step_number": (
"The index of the current step in the consensus workflow, beginning at 1. Step 1 is your analysis, " "The index of the current step in the consensus workflow, beginning at 1. Step 1 is your analysis, "
@@ -54,8 +55,11 @@ CONSENSUS_WORKFLOW_FIELD_DESCRIPTIONS = {
), ),
"next_step_required": ("Set to true if more models need to be consulted. False when ready for final synthesis."), "next_step_required": ("Set to true if more models need to be consulted. False when ready for final synthesis."),
"findings": ( "findings": (
"In step 1, provide your comprehensive analysis of the proposal. In steps 2+, summarize the key points " "In step 1: Provide YOUR OWN comprehensive analysis of the proposal/question. This is where you share your "
"from the model response received, noting agreements and disagreements with previous analyses." "independent evaluation, considering technical feasibility, risks, benefits, and alternatives. This analysis "
"is NOT sent to other models - it's recorded for the final synthesis. "
"In steps 2+: Summarize the key points from the model response received, noting agreements and disagreements "
"with previous analyses."
), ),
"relevant_files": ( "relevant_files": (
"Files that are relevant to the consensus analysis. Include files that help understand the proposal, " "Files that are relevant to the consensus analysis. Include files that help understand the proposal, "
@@ -161,6 +165,7 @@ class ConsensusTool(WorkflowTool):
def __init__(self): def __init__(self):
super().__init__() super().__init__()
self.initial_prompt: str | None = None self.initial_prompt: str | None = None
self.original_proposal: str | None = None # Store the original proposal separately
self.models_to_consult: list[dict] = [] self.models_to_consult: list[dict] = []
self.accumulated_responses: list[dict] = [] self.accumulated_responses: list[dict] = []
self._current_arguments: dict[str, Any] = {} self._current_arguments: dict[str, Any] = {}
@@ -394,7 +399,7 @@ of the evidence, even when it strongly points in one direction.""",
# Prepare final synthesis data # Prepare final synthesis data
response_data["complete_consensus"] = { response_data["complete_consensus"] = {
"initial_prompt": self.initial_prompt, "initial_prompt": self.original_proposal if self.original_proposal else self.initial_prompt,
"models_consulted": [m["model"] + ":" + m.get("stance", "neutral") for m in self.accumulated_responses], "models_consulted": [m["model"] + ":" + m.get("stance", "neutral") for m in self.accumulated_responses],
"total_responses": len(self.accumulated_responses), "total_responses": len(self.accumulated_responses),
"consensus_confidence": "high", # Consensus complete "consensus_confidence": "high", # Consensus complete
@@ -445,7 +450,9 @@ of the evidence, even when it strongly points in one direction.""",
# On first step, store the models to consult # On first step, store the models to consult
if request.step_number == 1: if request.step_number == 1:
self.initial_prompt = request.step # Store the original proposal from step 1 - this is what all models should see
self.original_proposal = request.step
self.initial_prompt = request.step # Keep for backward compatibility
self.models_to_consult = request.models or [] self.models_to_consult = request.models or []
self.accumulated_responses = [] self.accumulated_responses = []
# Set total steps: len(models) (each step includes consultation + response) # Set total steps: len(models) (each step includes consultation + response)
@@ -488,7 +495,7 @@ of the evidence, even when it strongly points in one direction.""",
response_data["status"] = "consensus_workflow_complete" response_data["status"] = "consensus_workflow_complete"
response_data["consensus_complete"] = True response_data["consensus_complete"] = True
response_data["complete_consensus"] = { response_data["complete_consensus"] = {
"initial_prompt": self.initial_prompt, "initial_prompt": self.original_proposal if self.original_proposal else self.initial_prompt,
"models_consulted": [ "models_consulted": [
f"{m['model']}:{m.get('stance', 'neutral')}" for m in self.accumulated_responses f"{m['model']}:{m.get('stance', 'neutral')}" for m in self.accumulated_responses
], ],
@@ -539,7 +546,9 @@ of the evidence, even when it strongly points in one direction.""",
# Prepare the prompt with any relevant files # Prepare the prompt with any relevant files
# Use continuation_id=None for blinded consensus - each model should only see # Use continuation_id=None for blinded consensus - each model should only see
# original prompt + files, not conversation history or other model responses # original prompt + files, not conversation history or other model responses
prompt = self.initial_prompt # CRITICAL: Use the original proposal from step 1, NOT what's in request.step for steps 2+!
# Steps 2+ contain summaries/notes that must NEVER be sent to other models
prompt = self.original_proposal if self.original_proposal else self.initial_prompt
if request.relevant_files: if request.relevant_files:
file_content, _ = self._prepare_file_content_for_prompt( file_content, _ = self._prepare_file_content_for_prompt(
request.relevant_files, request.relevant_files,
@@ -761,7 +770,8 @@ of the evidence, even when it strongly points in one direction.""",
def store_initial_issue(self, step_description: str): def store_initial_issue(self, step_description: str):
"""Store initial prompt for model consultations.""" """Store initial prompt for model consultations."""
self.initial_prompt = step_description self.original_proposal = step_description
self.initial_prompt = step_description # Keep for backward compatibility
# Required abstract methods from BaseTool # Required abstract methods from BaseTool
def get_request_model(self): def get_request_model(self):

6
tools/tracer.py
View File

@@ -535,9 +535,9 @@ class TracerTool(WorkflowTool):
tool_name = self.get_name() tool_name = self.get_name()
status_mapping = { status_mapping = {
f"{tool_name}_in_progress": "tracing_in_progress", f"{tool_name}_in_progress": "tracing_in_progress",
f"pause_for_{tool_name}": f"pause_for_tracing", f"pause_for_{tool_name}": "pause_for_tracing",
f"{tool_name}_required": f"tracing_required", f"{tool_name}_required": "tracing_required",
f"{tool_name}_complete": f"tracing_complete", f"{tool_name}_complete": "tracing_complete",
} }
if response_data["status"] in status_mapping: if response_data["status"] in status_mapping: