Chore/minimal cleanup

.env.example

@@ -1,36 +1,36 @@
-# Copy this file to `.env` for local development.
-# Do not commit real credentials.
-
-# Core LLM toggle
-CHAINCOPILOT_LLM_ENABLED=true
-CHAINCOPILOT_LLM_TIMEOUT_S=60
-CHAINCOPILOT_LLM_RETRY_ATTEMPTS=1
-CHAINCOPILOT_LLM_MODEL=gemma-4-31B-it
-
-# Planner mode
-# - hybrid: use LLM for candidate drafts, then deterministic scoring/selection
-# - deterministic: skip planner LLM calls entirely
-CHAINCOPILOT_PLANNER_MODE=hybrid
-
-# Provider selection
-# - gemini: Gemini Developer API
-# - vertex: Vertex AI Gemini via aiplatform.googleapis.com
-CHAINCOPILOT_LLM_PROVIDER=fpt
-
-# Gemini Developer API
-CHAINCOPILOT_LLM_API_KEY=
-
-# Vertex AI Gemini
-# Switch CHAINCOPILOT_LLM_PROVIDER=vertex when these are set correctly.
-# VERTEX_AI_API_KEY=REPLACE_WITH_VERTEX_API_KEY
-# VERTEX_AI_PROJECT_ID=REPLACE_WITH_GCP_PROJECT_ID
-# VERTEX_AI_REGION=global
-
-# Optional fallbacks recognized by the runtime
-# GEMINI_API_KEY=REPLACE_WITH_GEMINI_API_KEY
-# GOOGLE_CLOUD_PROJECT=REPLACE_WITH_GCP_PROJECT_ID
-# GOOGLE_CLOUD_LOCATION=global
-
-
-
-
+# Copy this file to `.env` for local development.
+# Do not commit real credentials.
+
+# Core LLM toggle
+CHAINCOPILOT_LLM_ENABLED=true
+CHAINCOPILOT_LLM_TIMEOUT_S=60
+CHAINCOPILOT_LLM_RETRY_ATTEMPTS=1
+CHAINCOPILOT_LLM_MODEL=gemma-4-31B-it
+
+# Planner mode
+# - hybrid: use LLM for candidate drafts, then deterministic scoring/selection
+# - deterministic: skip planner LLM calls entirely
+CHAINCOPILOT_PLANNER_MODE=hybrid
+
+# Provider selection
+# - gemini: Gemini Developer API
+# - vertex: Vertex AI Gemini via aiplatform.googleapis.com
+CHAINCOPILOT_LLM_PROVIDER=fpt
+
+# Gemini Developer API
+CHAINCOPILOT_LLM_API_KEY=
+
+# Vertex AI Gemini
+# Switch CHAINCOPILOT_LLM_PROVIDER=vertex when these are set correctly.
+# VERTEX_AI_API_KEY=REPLACE_WITH_VERTEX_API_KEY
+# VERTEX_AI_PROJECT_ID=REPLACE_WITH_GCP_PROJECT_ID
+# VERTEX_AI_REGION=global
+
+# Optional fallbacks recognized by the runtime
+# GEMINI_API_KEY=REPLACE_WITH_GEMINI_API_KEY
+# GOOGLE_CLOUD_PROJECT=REPLACE_WITH_GCP_PROJECT_ID
+# GOOGLE_CLOUD_LOCATION=global
+
+
+
+

.streamlit/config.toml

@@ -1,10 +1,10 @@
-[theme]
-primaryColor = "#4F46E5"
-backgroundColor = "#FFFFFF"
-secondaryBackgroundColor = "#F8FAFC"
-textColor = "#1E293B"
-font = "sans serif"
-base = "light"
-
-[server]
-headless = true
+[theme]
+primaryColor = "#4F46E5"
+backgroundColor = "#FFFFFF"
+secondaryBackgroundColor = "#F8FAFC"
+textColor = "#1E293B"
+font = "sans serif"
+base = "light"
+
+[server]
+headless = true

README.md

@@ -1,42 +1,42 @@
-# ChainCopilot
-
-ChainCopilot is a hackathon MVP for an autonomous resilient supply chain control tower.
-It implements a stateful agentic loop:
-
-- Sense
-- Analyze
-- Plan
-- Act
-- Learn
-
-The MVP includes:
-
-- 6-layer control-tower architecture
-- Demand, Inventory, Supplier, Logistics, Risk, and Planner agents
-- Normal and crisis modes with deterministic policy switching
-- Decision logs with explainability
-- What-if scenario simulation
-- Streamlit dashboard and FastAPI service
-
-## Quickstart
-
-```bash
-python3 -m venv .venv
-source .venv/bin/activate
-pip install -r requirements.txt
-pytest -q
-streamlit run ui/dashboard.py
-```
-
-## API
-
-```bash
-python run_api.py
-```
-
-## Scenarios
-
-- supplier_delay
-- demand_spike
-- route_blockage
-- compound_disruption
+# ChainCopilot
+
+ChainCopilot is a hackathon MVP for an autonomous resilient supply chain control tower.
+It implements a stateful agentic loop:
+
+- Sense
+- Analyze
+- Plan
+- Act
+- Learn
+
+The MVP includes:
+
+- 6-layer control-tower architecture
+- Demand, Inventory, Supplier, Logistics, Risk, and Planner agents
+- Normal and crisis modes with deterministic policy switching
+- Decision logs with explainability
+- What-if scenario simulation
+- Streamlit dashboard and FastAPI service
+
+## Quickstart
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+pip install -r requirements.txt
+pytest -q
+streamlit run ui/dashboard.py
+```
+
+## API
+
+```bash
+python run_api.py
+```
+
+## Scenarios
+
+- supplier_delay
+- demand_spike
+- route_blockage
+- compound_disruption

actions/switch_supplier.py

@@ -1,26 +1,26 @@
-from core.models import Action, SystemState
-
-
-def apply_switch_supplier(state: SystemState, action: Action) -> None:
-    item = state.inventory.get(action.target_id)
-    if item is None:
-        return
-
-    supplier_id = action.parameters.get("supplier_id")
-    if not supplier_id:
-        return
-
-    supplier = _resolve_supplier(state, supplier_id, item.sku)
-    if supplier is None:
-        return
-
-    item.preferred_supplier_id = supplier_id
-    item.unit_cost = supplier.unit_cost
-    state.extra_cost += max(action.estimated_cost_delta, 0.0)
-
-
-def _resolve_supplier(state: SystemState, supplier_id: str, sku: str):
-    return (
-        state.suppliers.get(f"{supplier_id}_{sku}")
-        or state.suppliers.get(supplier_id)
-    )
+from core.models import Action, SystemState
+
+
+def apply_switch_supplier(state: SystemState, action: Action) -> None:
+    item = state.inventory.get(action.target_id)
+    if item is None:
+        return
+
+    supplier_id = action.parameters.get("supplier_id")
+    if not supplier_id:
+        return
+
+    supplier = _resolve_supplier(state, supplier_id, item.sku)
+    if supplier is None:
+        return
+
+    item.preferred_supplier_id = supplier_id
+    item.unit_cost = supplier.unit_cost
+    state.extra_cost += max(action.estimated_cost_delta, 0.0)
+
+
+def _resolve_supplier(state: SystemState, supplier_id: str, sku: str):
+    return (
+        state.suppliers.get(f"{supplier_id}_{sku}")
+        or state.suppliers.get(supplier_id)
+    )

agents/base.py

@@ -1,34 +1,34 @@
-from __future__ import annotations
-
-from abc import ABC, abstractmethod
-
-from core.models import AgentProposal, Event, SystemState
-
-
-class BaseAgent(ABC):
-    name: str
-    custom_system_prompt: str | None = None
-
-    @abstractmethod
-    def run(self, state: SystemState, event: Event | None = None) -> AgentProposal:
-        raise NotImplementedError
-
-    def enrich_with_llm(
-        self,
-        *,
-        state: SystemState,
-        event: Event | None,
-        proposal: AgentProposal,
-        state_slice: dict,
-    ) -> AgentProposal:
-        from llm.service import enrich_specialist_proposal
-
-        enrich_specialist_proposal(
-            agent_name=self.name,
-            state=state,
-            event=event,
-            proposal=proposal,
-            state_slice=state_slice,
-            custom_prompt=self.custom_system_prompt,
-        )
-        return proposal
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+
+from core.models import AgentProposal, Event, SystemState
+
+
+class BaseAgent(ABC):
+    name: str
+    custom_system_prompt: str | None = None
+
+    @abstractmethod
+    def run(self, state: SystemState, event: Event | None = None) -> AgentProposal:
+        raise NotImplementedError
+
+    def enrich_with_llm(
+        self,
+        *,
+        state: SystemState,
+        event: Event | None,
+        proposal: AgentProposal,
+        state_slice: dict,
+    ) -> AgentProposal:
+        from llm.service import enrich_specialist_proposal
+
+        enrich_specialist_proposal(
+            agent_name=self.name,
+            state=state,
+            event=event,
+            proposal=proposal,
+            state_slice=state_slice,
+            custom_prompt=self.custom_system_prompt,
+        )
+        return proposal

agents/critic.py

@@ -3,6 +3,7 @@
 from agents.base import BaseAgent
 from core.models import AgentProposal, Event, SystemState
 from llm.service import critique_candidate_plans
+from policies.explainability import build_critic_review
 
 
 class CriticAgent(BaseAgent):
@@ -21,6 +22,11 @@ def run(self, state: SystemState, event: Event | None = None) -> AgentProposal:
             selected_plan=state.latest_plan,
             evaluations=decision_log.candidate_evaluations,
         )
+        if not summary and not findings:
+            summary, findings = build_critic_review(
+                state.latest_plan,
+                decision_log.candidate_evaluations,
+            )
         state.latest_plan.critic_summary = summary
         decision_log.critic_summary = summary
         decision_log.critic_findings = findings
@@ -32,11 +38,12 @@ def run(self, state: SystemState, event: Event | None = None) -> AgentProposal:
             proposal.notes_for_planner = summary
         if findings:
             proposal.downstream_impacts = findings
+            proposal.tradeoffs = findings[:2]
         proposal.llm_used = used
         proposal.llm_error = error
         proposal.observations.append(
-            "critic reviewed candidate plans"
+            "Critic reviewed candidate plans"
             if used or summary or findings
-            else "critic unavailable; deterministic selection retained"
+            else "Critic unavailable; deterministic selection retained"
         )
         return proposal