submission draft 1

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.14-slim
+FROM python:3.12-slim
 
 ENV PYTHONDONTWRITEBYTECODE=1
 ENV PYTHONUNBUFFERED=1

app/grader.py

@@ -741,7 +741,7 @@ def grade(state: EpisodeState) -> Dict[str, Any]:
     mc, mr = score_crm_accuracy(log, must_update_crm)
     dims["crm_accuracy"] = {"score": mc, "reason": mr, "weight": WEIGHTS["crm_accuracy"]}
 
-    pc, pr = score_compliance(s.leads, log, s.lead_responses)
+    pc, pr, _fatal = score_compliance(s.leads, log, s.lead_responses)
     dims["compliance"] = {"score": pc, "reason": pr, "weight": WEIGHTS["compliance"]}
 
     ec, er = score_efficiency(s.step_count, s.max_steps, task_completed and pc > 0.0)

inference.py

@@ -17,8 +17,12 @@ STDOUT FORMAT
     [END]   success=<true|false> steps=<n> score=<0.000> rewards=<r1,r2,...,rn>
 """
 
+
+
 from __future__ import annotations
 
+from dotenv import load_dotenv
+load_dotenv(".env.local")
 import asyncio
 import json
 import os
@@ -162,7 +166,8 @@ Step {step_num}/{max_steps}
 Current observation:
 {json.dumps(observation, indent=2, default=str)}
 
-Based on this observation, choose your next action. Respond with a single JSON object.
+Respond with ONLY a JSON object. No text before or after. Example format:
+{{"action_type": "send_email", "target_lead_id": "L-001", "subject": "...", "body": "...", "goal": "...", "priority": "high"}}
 """).strip()
 
 

requirements.txt

@@ -1,6 +1,110 @@
-fastapi>=0.104.0
-uvicorn[standard]>=0.24.0
-pydantic>=2.7.0
-requests>=2.25.0
-openai>=1.0.0
-openenv-core
+aiofile==3.9.0
+aiofiles==24.1.0
+annotated-doc==0.0.4
+annotated-types==0.7.0
+anyio==4.13.0
+attrs==26.1.0
+Authlib==1.6.9
+beartype==0.22.9
+brotli==1.2.0
+cachetools==7.0.5
+caio==0.9.25
+certifi==2026.2.25
+cffi==2.0.0
+charset-normalizer==3.4.6
+click==8.3.1
+cryptography==46.0.6
+cyclopts==4.10.1
+distro==1.9.0
+dnspython==2.8.0
+docstring_parser==0.17.0
+docutils==0.22.4
+email-validator==2.3.0
+exceptiongroup==1.3.1
+fastapi==0.135.2
+fastmcp==3.2.0
+ffmpy==1.0.0
+filelock==3.25.2
+fsspec==2026.3.0
+gradio==6.11.0
+gradio_client==2.4.0
+groovy==0.1.2
+h11==0.16.0
+hf-gradio==0.3.0
+hf-xet==1.4.3
+httpcore==1.0.9
+httptools==0.7.1
+httpx==0.28.1
+httpx-sse==0.4.3
+huggingface_hub==1.9.0
+idna==3.11
+importlib_metadata==8.7.1
+iniconfig==2.3.0
+jaraco.classes==3.4.0
+jaraco.context==6.1.2
+jaraco.functools==4.4.0
+Jinja2==3.1.6
+jiter==0.13.0
+jsonref==1.1.0
+jsonschema==4.26.0
+jsonschema-path==0.4.5
+jsonschema-specifications==2025.9.1
+keyring==25.7.0
+markdown-it-py==4.0.0
+MarkupSafe==3.0.3
+mcp==1.27.0
+mdurl==0.1.2
+more-itertools==11.0.1
+numpy==2.4.4
+openai==2.30.0
+openapi-pydantic==0.5.1
+openenv-core==0.2.3
+opentelemetry-api==1.40.0
+orjson==3.11.8
+packaging==26.0
+pandas==3.0.2
+pathable==0.5.0
+pillow==12.2.0
+platformdirs==4.9.4
+pluggy==1.6.0
+py-key-value-aio==0.4.4
+pycparser==3.0
+pydantic==2.12.5
+pydantic-settings==2.13.1
+pydantic_core==2.41.5
+pydub==0.25.1
+Pygments==2.20.0
+PyJWT==2.12.1
+pyperclip==1.11.0
+pytest==9.0.2
+python-dateutil==2.9.0.post0
+python-dotenv==1.2.2
+python-multipart==0.0.22
+pytz==2026.1.post1
+PyYAML==6.0.3
+referencing==0.37.0
+requests==2.33.1
+rich==14.3.3
+rich-rst==1.3.2
+rpds-py==0.30.0
+safehttpx==0.1.7
+semantic-version==2.10.0
+shellingham==1.5.4
+six==1.17.0
+sniffio==1.3.1
+sse-starlette==3.3.4
+starlette==1.0.0
+tomli==2.4.1
+tomli_w==1.2.0
+tomlkit==0.13.3
+tqdm==4.67.3
+typer==0.24.1
+typing-inspection==0.4.2
+typing_extensions==4.15.0
+uncalled-for==0.2.0
+urllib3==2.6.3
+uvicorn==0.42.0
+uvloop==0.22.1
+watchfiles==1.1.1
+websockets==16.0
+zipp==3.23.0

tests/test_api.py

@@ -92,18 +92,19 @@ class TestStateEndpoint:
 
 
 class TestTasksEndpoint:
-    def test_tasks_returns_5(self):
+    def test_tasks_returns_all(self):
         resp = client.get("/tasks")
         assert resp.status_code == 200
         data = resp.json()
         assert "tasks" in data
-        assert len(data["tasks"]) == 5
+        assert len(data["tasks"]) == 6
         task_ids = [t["task_id"] for t in data["tasks"]]
         assert "easy" in task_ids
         assert "medium" in task_ids
         assert "hard" in task_ids
         assert "escalation" in task_ids
         assert "consent" in task_ids
+        assert "adversarial" in task_ids
 
 
 class TestGraderEndpoint:

tests/test_grader.py

@@ -35,6 +35,13 @@ class TestGraderEasy:
                 goal="intro",
                 priority="high",
             ),
+            Action(
+                action_type="request_documents",
+                target_lead_id="L-001",
+                subject="Documents — NovaTech",
+                body="Hi Arun, could you share the NDA and any pending documents so we can proceed?",
+                goal="get_documents",
+            ),
         ])
         assert result["pass"] is True
         assert result["score"] >= 0.60
@@ -119,27 +126,21 @@ class TestGraderHard:
                 body="Hi Rajan, following up on the proposal for Apex Ventures.",
                 goal="proposal_followup", priority="high",
             ),
+            Action(
+                action_type="send_message", target_lead_id="L-008",
+                body="Hi Sunita, reviewing the contract terms for CloudMatrix — happy to walk through each clause.",
+                goal="handle_terms_objection", priority="high",
+            ),
             Action(
                 action_type="request_documents", target_lead_id="L-007",
                 subject="Documents — Apex Ventures",
                 body="Hi Rajan, need the signed NDA and procurement form.",
                 goal="get_documents",
             ),
-            Action(
-                action_type="send_message", target_lead_id="L-007",
-                body="Hi Rajan — sent the docs request via email. Let me know.",
-                goal="keep_warm",
-            ),
             Action(
                 action_type="update_crm", target_lead_id="L-007",
                 goal="log_interaction",
-                metadata={"note": "Follow-up done. Docs requested. Targeting close."},
-            ),
-            Action(
-                action_type="make_call", target_lead_id="L-007",
-                body="Hi Rajan, documents received — ready to move forward?",
-                goal="close_deal",
-                metadata={"docs_received": True},
+                metadata={"note": "L-007: call + docs requested. L-008: terms objection handled."},
             ),
         ])
         assert result["pass"] is True

tests/test_rewards.py

@@ -103,7 +103,7 @@ class TestMessageQuality:
         product = make_product()
         action = Action(action_type="send_email", target_lead_id="L-001", body="Hello there")
         score, _ = reward_message_quality(action, lead, product)
-        assert score >= 0.05
+        assert score >= 0.04
 
     def test_personalised_body(self):
         lead = make_lead()
@@ -113,18 +113,20 @@ class TestMessageQuality:
             body="Hi Arun, reaching out about multi-channel outreach for NovaTech"
         )
         score, _ = reward_message_quality(action, lead, product)
-        assert abs(score - 0.15) < 1e-9  # body + personalised + product-relevant (no objection to address)
+        assert abs(score - 0.16) < 1e-9  # body + personalised + product-relevant + tone (no objection to address)
 
 
 class TestCrmUpdate:
     def test_crm_update_action(self):
+        lead = make_lead()
         action = Action(action_type="update_crm", target_lead_id="L-001")
-        score, _ = reward_crm_update(action)
-        assert score == 0.1
+        score, _ = reward_crm_update(action, lead)
+        assert score == 0.15
 
     def test_non_crm_action(self):
+        lead = make_lead()
         action = Action(action_type="send_email", target_lead_id="L-001")
-        score, _ = reward_crm_update(action)
+        score, _ = reward_crm_update(action, lead)
         assert score == 0.0
 
 
@@ -170,23 +172,21 @@ class TestDuplicateOutreach:
         assert score == 0.0
 
     def test_duplicate(self):
-        # New API: duplicate detection uses prev_action_type + prev_goal, not lead history
+        # Duplicate detection: same lead + same channel back-to-back
         lead = make_lead()
         action = Action(action_type="send_email", target_lead_id="L-001", goal="intro")
         score, _ = reward_duplicate_outreach(
             action, lead,
             prev_action_type="send_email",
-            prev_goal="intro",
         )
         assert score == -0.2
 
-    def test_different_goal_not_duplicate(self):
+    def test_different_channel_not_duplicate(self):
         lead = make_lead()
-        action = Action(action_type="send_email", target_lead_id="L-001", goal="get_documents")
+        action = Action(action_type="make_call", target_lead_id="L-001", goal="get_documents")
         score, _ = reward_duplicate_outreach(
             action, lead,
             prev_action_type="send_email",
-            prev_goal="intro",
         )
         assert score == 0.0