codee-agent 0.6.5__tar.gz → 0.6.7__tar.gz

{codee_agent-0.6.5 → codee_agent-0.6.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codee-agent
-Version: 0.6.5
+Version: 0.6.7
 Summary: A virtual co-worker that picks up Jira and Azure DevOps tasks and solves them with coding agents.
 Keywords: ai,agent,jira,azure-devops,claude-code,github-copilot,codex,automation
 Author: Denis Kibalko

{codee_agent-0.6.5 → codee_agent-0.6.7}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "codee-agent"
-version = "0.6.5"
+version = "0.6.7"
 description = "A virtual co-worker that picks up Jira and Azure DevOps tasks and solves them with coding agents."
 readme = "README.md"
 license = "MIT"

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee/admin.py

@@ -273,6 +273,7 @@ class RunRecord(BaseModel):
     message: str
     user_message: str
     response: str
+    debug_logs: str
     preview: str
     viewer_url: str
 
@@ -367,6 +368,10 @@ class AdminState(rx.State):
     agent_test_messages: list[ConversationMessage] = []
     agent_test_sending: bool = False
     agent_test_generation: int = 0
+    agent_test_model: str = ""
+    agent_test_models: list[ModelOption] = []
+    agent_test_model_query: str = ""
+    agent_test_models_loading: bool = False
     # Whether the executor runs Claude Code on the connected accounts instead
     # of leaving ~/.claude/.credentials.json alone.
     claude_code_rotate_keys: bool = False
@@ -475,6 +480,32 @@ class AdminState(rx.State):
             return ""
         return query
 
+    @rx.var
+    def filtered_agent_test_models(self) -> list[ModelOption]:
+        query = self.agent_test_model_query.strip().lower()
+        return [
+            model for model in self.agent_test_models
+            if not query or query in f"{model.name} {model.id}".lower()
+        ]
+
+    @rx.var
+    def agent_test_model_label(self) -> str:
+        if not self.agent_test_model:
+            return "Agent default"
+        for model in self.agent_test_models:
+            if model.id == self.agent_test_model:
+                return model.name
+        return self.agent_test_model
+
+    @rx.var
+    def custom_agent_test_model_query(self) -> str:
+        query = self.agent_test_model_query.strip()
+        if not query or any(
+            model.id == query for model in self.agent_test_models
+        ):
+            return ""
+        return query
+
     @rx.var
     def active_route(self) -> str:
         return self.router.url.path.rstrip("/") or "/"
@@ -826,6 +857,7 @@ class AdminState(rx.State):
                 message=message,
                 user_message=(run.get("user_message") or message).strip(),
                 response=(run.get("response") or "").strip(),
+                debug_logs=run.get("debug_logs") or "",
                 preview=preview[:120] + ("..." if len(preview) > 120 else ""),
                 viewer_url=(SERVICE.session_viewer.format(session_id=run["session_id"])
                             if SERVICE.session_viewer and run.get("session_id") else ""),
@@ -1527,13 +1559,17 @@ class AdminState(rx.State):
         error = self._persist_settings()
         return rx.toast.error(error) if error else rx.toast.success("Settings saved")
 
-    def open_agent_test(self) -> None:
+    def open_agent_test(self) -> Any:
         self.agent_test_input = ""
         self.agent_test_session_id = ""
         self.agent_test_messages = []
         self.agent_test_sending = False
         self.agent_test_generation += 1
+        self.agent_test_model = ""
+        self.agent_test_models = []
+        self.agent_test_model_query = ""
         self.agent_test_open = True
+        return AdminState.load_agent_test_models
 
     def set_agent_test_open(self, open_: bool) -> None:
         self.agent_test_open = open_
@@ -1544,6 +1580,28 @@ class AdminState(rx.State):
     def set_agent_test_input(self, value: str) -> None:
         self.agent_test_input = value
 
+    def set_agent_test_model_query(self, value: str) -> None:
+        self.agent_test_model_query = value
+
+    def choose_agent_test_model(self, model_id: str) -> None:
+        self.agent_test_model = model_id.strip()
+        self.agent_test_model_query = ""
+
+    @rx.event(background=True)
+    async def load_agent_test_models(self) -> None:
+        async with self:
+            agent = self.coding_agent
+            self.agent_test_models_loading = True
+        try:
+            models = await asyncio.to_thread(SERVICE.list_agent_models, agent)
+        except Exception:
+            models = []
+        async with self:
+            if self.coding_agent != agent:
+                return
+            self.agent_test_models = [ModelOption(**model) for model in models]
+            self.agent_test_models_loading = False
+
     @rx.event(background=True)
     async def send_agent_test_message(self) -> Any:
         async with self:
@@ -1552,6 +1610,7 @@ class AdminState(rx.State):
                 return
             agent = self.coding_agent
             session_id = self.agent_test_session_id
+            model = self.agent_test_model
             generation = self.agent_test_generation
             self.agent_test_input = ""
             self.agent_test_sending = True
@@ -1565,6 +1624,7 @@ class AdminState(rx.State):
                 agent,
                 message,
                 session_id,
+                model,
             )
         except Exception as error:
             async with self:
@@ -2131,7 +2191,7 @@ def model_menu_item(button: rx.Component) -> rx.Component:
     return rx.popover.close(rx.flex(button, width="100%"), width="100%")
 
 
-def model_option_row(option: ModelOption) -> rx.Component:
+def _model_option_row(option: ModelOption, choose_model: Any) -> rx.Component:
     """One row of the model picker: friendly name left, model code right."""
     return model_menu_item(
         rx.button(
@@ -2143,7 +2203,15 @@ def model_option_row(option: ModelOption) -> rx.Component:
                       align="center", spacing="2", width="100%"),
             variant="ghost", color_scheme="gray", width="100%",
             justify_content="start", padding="0.45rem 0.6rem",
-            on_click=AdminState.choose_model(option.id)))
+            on_click=choose_model(option.id)))
+
+
+def model_option_row(option: ModelOption) -> rx.Component:
+    return _model_option_row(option, AdminState.choose_model)
+
+
+def agent_test_model_option_row(option: ModelOption) -> rx.Component:
+    return _model_option_row(option, AdminState.choose_agent_test_model)
 
 
 def agent_picker() -> rx.Component:
@@ -2481,7 +2549,7 @@ def run_row(run: RunRecord) -> rx.Component:
             rx.cond(run.viewer_url != "", rx.link(rx.icon("external-link", size=16), href=run.viewer_url,
                                                   is_external=True, aria_label="View session", color=ACCENT)),
             gap="1rem", align="start", width="100%"),
-        rx.cond((run.user_message != "") | (run.response != ""), rx.accordion.root(rx.accordion.item(
+        rx.cond((run.user_message != "") | (run.response != "") | (run.debug_logs != ""), rx.accordion.root(rx.accordion.item(
             header="Run info", content=rx.vstack(
                 rx.text("User message", font_weight="600"),
                 rx.text(run.user_message, white_space="pre-wrap"),
@@ -2489,6 +2557,12 @@ def run_row(run: RunRecord) -> rx.Component:
                     rx.text("LLM response", font_weight="600",
                             margin_top="0.75rem"),
                     rx.text(run.response, white_space="pre-wrap"))),
+                rx.cond(run.debug_logs != "", rx.fragment(
+                        rx.text("Debug logs", font_weight="600",
+                                margin_top="0.75rem"),
+                        rx.text(run.debug_logs, white_space="pre-wrap",
+                                font_family="IBM Plex Mono, monospace",
+                                font_size="0.8rem"))),
                 spacing="2", align="start", width="100%"), value=run.started_at),
             collapsible=True, width="100%")),
         padding="1rem", background=SURFACE, border=BORDER, width="100%")
@@ -3309,6 +3383,72 @@ def agent_test_dialog() -> rx.Component:
             rx.dialog.description(
                 "Chat with the selected coding agent in the Codee project.",
                 color=MUTED),
+            field(
+                "Model",
+                rx.popover.root(
+                    rx.popover.trigger(
+                        rx.button(
+                            rx.hstack(
+                                rx.text(AdminState.agent_test_model_label),
+                                rx.spacer(),
+                                rx.icon("chevrons-up-down", size=14),
+                                align="center", width="100%"),
+                            variant="surface", color_scheme="gray",
+                            width="100%", type="button")),
+                    rx.popover.content(
+                        rx.vstack(
+                            rx.input(
+                                placeholder=(
+                                    "Search models, or type a model code"),
+                                value=AdminState.agent_test_model_query,
+                                on_change=AdminState.set_agent_test_model_query,
+                                auto_focus=True, width="100%"),
+                            rx.cond(
+                                AdminState.custom_agent_test_model_query != "",
+                                model_menu_item(
+                                    rx.button(
+                                        rx.hstack(
+                                            rx.icon("plus", size=14),
+                                            rx.text("Use "),
+                                            rx.code(
+                                                AdminState.custom_agent_test_model_query),
+                                            align="center", spacing="2"),
+                                        variant="soft", width="100%",
+                                        justify_content="start",
+                                        padding="0.45rem 0.6rem",
+                                        on_click=AdminState.choose_agent_test_model(
+                                            AdminState.custom_agent_test_model_query)))),
+                            rx.scroll_area(
+                                rx.vstack(
+                                    model_menu_item(
+                                        rx.button(
+                                            "Agent default", variant="ghost",
+                                            color_scheme="gray", width="100%",
+                                            justify_content="start",
+                                            padding="0.45rem 0.6rem",
+                                            on_click=AdminState.choose_agent_test_model(""))),
+                                    rx.foreach(
+                                        AdminState.filtered_agent_test_models,
+                                        agent_test_model_option_row),
+                                    rx.cond(
+                                        AdminState.agent_test_models_loading,
+                                        rx.text(
+                                            "Loading models from the coding agent…",
+                                            color=MUTED, font_size="0.8rem",
+                                            padding="0.5rem")),
+                                    spacing="1", width="100%"),
+                                type="auto", scrollbars="vertical",
+                                max_height="15rem", width="100%"),
+                            spacing="2", width="100%"),
+                        width="24rem", max_width="calc(100vw - 3rem)")),
+                rx.text(
+                    rx.cond(
+                        AdminState.agent_test_model == "",
+                        "Runs on whatever that agent defaults to.",
+                        rx.fragment("Uses ",
+                                    rx.code(AdminState.agent_test_model),
+                                    " for this conversation.")),
+                    color=MUTED, font_size="0.82rem")),
             rx.scroll_area(
                 rx.vstack(
                     rx.cond(
@@ -3374,7 +3514,7 @@ def settings_page() -> rx.Component:
             rx.heading("Coding agent", size="4", margin_bottom="1rem"),
             rx.vstack(
                 field("Default agent",
-                      rx.hstack(
+                      rx.grid(
                           rx.select(
                               ["claude_code", "github_copilot", "codex"],
                               value=AdminState.coding_agent,
@@ -3384,7 +3524,10 @@ def settings_page() -> rx.Component:
                                     "Test conversation", variant="outline",
                                     white_space="nowrap",
                                     on_click=AdminState.open_agent_test),
-                          width="100%", spacing="3")),
+                          grid_template_columns=rx.breakpoints(
+                              initial="minmax(0, 1fr)",
+                              md="minmax(0, 1fr) auto"),
+                          width="100%", gap="0.75rem")),
                 field("Max parallel tasks",
                       rx.input(value=AdminState.max_parallel_agents,
                                on_change=AdminState.set_max_parallel_agents,

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee/admin_service.py

@@ -2444,7 +2444,8 @@ class AdminService:
             raise RuntimeError(str(exc)) from exc
 
     def test_agent_conversation(
-        self, agent_code: str, user_message: str, session_id: str = ""
+        self, agent_code: str, user_message: str, session_id: str = "",
+        model: str = "",
     ) -> tuple[str, str]:
         """Run one turn with the selected agent and return reply plus session id."""
         selected = resolve_agent_code(agent_code)
@@ -2459,10 +2460,10 @@ class AdminService:
 
         if session_id:
             response = agent.continue_conversation(
-                user_message, session_id, on_session_id=opened)
+                user_message, session_id, model, on_session_id=opened)
         else:
             response = agent.run(
-                user_message, actual_session_id, on_session_id=opened)
+                user_message, actual_session_id, model, on_session_id=opened)
         return response, actual_session_id
 
     def setup_tasks_mcp(

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee/lib/runs_db.py

@@ -11,7 +11,7 @@ from codee_main_context.context import CodeeMainContext
 from codee_database.database import get_db_connection
 
 _COLUMNS = ("id", "skill_name", "trigger_type", "session_id", "status", "error",
-            "started_at", "message", "user_message", "response")
+            "started_at", "message", "user_message", "response", "debug_logs")
 
 # Codee names a session before the agent runs, but not every agent runs under
 # the name it was given: Codex mints its own thread id and reports it back mid
@@ -52,7 +52,8 @@ def init(main_context: CodeeMainContext) -> None:
                 started_at TEXT NOT NULL,
                 message TEXT,
                 user_message TEXT,
-                response TEXT
+                response TEXT,
+                debug_logs TEXT
             )"""
         )
         conn.execute(
@@ -65,6 +66,8 @@ def init(main_context: CodeeMainContext) -> None:
             conn.execute("ALTER TABLE runs ADD COLUMN response TEXT")
         if "user_message" not in cols:
             conn.execute("ALTER TABLE runs ADD COLUMN user_message TEXT")
+        if "debug_logs" not in cols:
+            conn.execute("ALTER TABLE runs ADD COLUMN debug_logs TEXT")
         # In-flight claude runs; a row lives only while its subprocess is running.
         conn.execute(
             """CREATE TABLE IF NOT EXISTS active_jobs (
@@ -77,28 +80,33 @@ def init(main_context: CodeeMainContext) -> None:
             )"""
         )
         # Migrate DBs created before the dashboard named the agent behind a run.
-        job_cols = {row[1] for row in conn.execute("PRAGMA table_info(active_jobs)")}
+        job_cols = {row[1] for row in conn.execute(
+            "PRAGMA table_info(active_jobs)")}
         for column in ("agent", "model"):
             if column not in job_cols:
-                conn.execute(f"ALTER TABLE active_jobs ADD COLUMN {column} TEXT")
+                conn.execute(
+                    f"ALTER TABLE active_jobs ADD COLUMN {column} TEXT")
 
 
 def record_run(skill_name, trigger_type, session_id, status, error=None, started_at=None,
-               message=None, user_message=None, response=None,
+               message=None, user_message=None, response=None, debug_logs=None,
                *, main_context: CodeeMainContext) -> None:
     """Insert one run row. Never raises to the caller (FR-009)."""
     try:
         init(main_context)
         session_id = agent_session(session_id)
+        if debug_logs is None:
+            debug_logs = getattr(response, "debug_logs", None)
         if started_at is None:
             started_at = datetime.now(timezone.utc).isoformat()
         with get_db_connection(main_context) as conn:
             conn.execute(
                 "INSERT INTO runs (skill_name, trigger_type, session_id, status, error,"
-                " started_at, message, user_message, response)"
-                " VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)",
+                " started_at, message, user_message, response, debug_logs)"
+                " VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
                 (skill_name, trigger_type, session_id,
-                 status, error, started_at, message, user_message, response),
+                 status, error, started_at, message, user_message, response,
+                 debug_logs),
             )
     except Exception as exc:  # ponytail: a logging miss must never abort the skill run
         print(f"[runs_db] Failed to record run for {skill_name}: {exc}")
@@ -112,7 +120,7 @@ def recent_runs(limit: int = 100, offset: int = 0, *,
         with get_db_connection(main_context) as conn:
             rows = conn.execute(
                 "SELECT id, skill_name, trigger_type, session_id, status, error,"
-                " started_at, message, user_message, response"
+                " started_at, message, user_message, response, debug_logs"
                 " FROM runs ORDER BY started_at DESC, id DESC LIMIT ? OFFSET ?",
                 (limit, max(offset, 0)),
             ).fetchall()

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee/lib/test_runs_db.py

@@ -30,7 +30,8 @@ def test_round_trip_newest_first(tmp_path):
                        started_at="2026-06-27T11:00:00+00:00", main_context=ctx)
 
     rows = runs_db.recent_runs(main_context=ctx)
-    assert [r["skill_name"] for r in rows] == ["skill-b", "skill-a"]  # newest first
+    assert [r["skill_name"] for r in rows] == [
+        "skill-b", "skill-a"]  # newest first
     assert rows[0]["trigger_type"] == "email"
     assert rows[0]["status"] == "failed"
     assert rows[0]["error"] == "boom"
@@ -74,7 +75,8 @@ def test_main_context_is_required():
             call()
         except TypeError:
             continue
-        raise AssertionError("main_context must be a required keyword argument")
+        raise AssertionError(
+            "main_context must be a required keyword argument")
 
 
 # ---------------------------------------------------------------- counts() (US1)
@@ -85,9 +87,12 @@ def _ago(hours):
 def test_counts_total_and_last_24h(tmp_path):
     ctx = _ctx(tmp_path)
     # Two inside the 24h window, one well outside, one exactly at the boundary (excluded, strict >).
-    runs_db.record_run("a", "cron", "s1", "succeeded", started_at=_ago(1), main_context=ctx)
-    runs_db.record_run("b", "cron", "s2", "succeeded", started_at=_ago(23), main_context=ctx)
-    runs_db.record_run("c", "cron", "s3", "succeeded", started_at=_ago(48), main_context=ctx)
+    runs_db.record_run("a", "cron", "s1", "succeeded",
+                       started_at=_ago(1), main_context=ctx)
+    runs_db.record_run("b", "cron", "s2", "succeeded",
+                       started_at=_ago(23), main_context=ctx)
+    runs_db.record_run("c", "cron", "s3", "succeeded",
+                       started_at=_ago(48), main_context=ctx)
     assert runs_db.counts(ctx) == {"total": 3, "last_24h": 2}
 
 
@@ -101,17 +106,22 @@ def test_counts_never_raises_on_bad_path(tmp_path):
 
 def test_runs_by_hour_buckets(tmp_path):
     ctx = _ctx(tmp_path)
-    runs_db.record_run("a", "cron", "s1", "succeeded", started_at=_ago(2), main_context=ctx)
-    runs_db.record_run("b", "cron", "s2", "succeeded", started_at=_ago(2.1), main_context=ctx)
-    runs_db.record_run("c", "cron", "s3", "succeeded", started_at=_ago(48), main_context=ctx)  # out of window
+    runs_db.record_run("a", "cron", "s1", "succeeded",
+                       started_at=_ago(2), main_context=ctx)
+    runs_db.record_run("b", "cron", "s2", "succeeded",
+                       started_at=_ago(2.1), main_context=ctx)
+    runs_db.record_run("c", "cron", "s3", "succeeded",
+                       started_at=_ago(48), main_context=ctx)  # out of window
 
     hourly = runs_db.runs_by_hour(ctx)
     assert len(hourly) == 24  # always 24 buckets
     # oldest-first: buckets are the last 24 hour labels in ascending time order
     now = datetime.now(timezone.utc).replace(minute=0, second=0, microsecond=0)
-    expected = [(now - timedelta(hours=h)).strftime("%H:00") for h in range(23, -1, -1)]
+    expected = [(now - timedelta(hours=h)).strftime("%H:00")
+                for h in range(23, -1, -1)]
     assert [h["hour"] for h in hourly] == expected
-    assert sum(h["runs"] for h in hourly) == 2  # 48h-old run excluded; the two ~2h-old runs counted
+    # 48h-old run excluded; the two ~2h-old runs counted
+    assert sum(h["runs"] for h in hourly) == 2
 
 
 def test_runs_by_hour_empty_and_bad_path(tmp_path):
@@ -123,10 +133,14 @@ def test_runs_by_hour_empty_and_bad_path(tmp_path):
 # ---------------------------------------------------------------- message column (US2)
 def test_message_round_trip(tmp_path):
     ctx = _ctx(tmp_path)
-    runs_db.record_run("a", "cron", "s1", "succeeded", message="hello", main_context=ctx)
-    runs_db.record_run("b", "email", "s2", "succeeded", message="", main_context=ctx)
-    runs_db.record_run("c", "aws-sqs", "s3", "succeeded", main_context=ctx)  # message defaults None
-    by_skill = {r["skill_name"]: r["message"] for r in runs_db.recent_runs(main_context=ctx)}
+    runs_db.record_run("a", "cron", "s1", "succeeded",
+                       message="hello", main_context=ctx)
+    runs_db.record_run("b", "email", "s2", "succeeded",
+                       message="", main_context=ctx)
+    runs_db.record_run("c", "aws-sqs", "s3", "succeeded",
+                       main_context=ctx)  # message defaults None
+    by_skill = {r["skill_name"]: r["message"]
+                for r in runs_db.recent_runs(main_context=ctx)}
     assert by_skill == {"a": "hello", "b": "", "c": None}
 
 
@@ -141,6 +155,19 @@ def test_response_round_trip(tmp_path):
     assert run["response"] == "final answer"
 
 
+def test_debug_logs_are_extracted_from_the_agent_response(tmp_path):
+    from codee_agent_abstract.provider import AgentResponse
+
+    ctx = _ctx(tmp_path)
+    response = AgentResponse("final answer", "debug one\ndebug two\n")
+    runs_db.record_run("a", "cron", "s1", "succeeded", response=response,
+                       main_context=ctx)
+
+    run, = runs_db.recent_runs(main_context=ctx)
+    assert run["response"] == "final answer"
+    assert run["debug_logs"] == "debug one\ndebug two\n"
+
+
 # ---------------------------------------------------------------- active_jobs
 def test_active_job_lifecycle(tmp_path):
     ctx = _ctx(tmp_path)
@@ -195,7 +222,8 @@ def test_active_jobs_elapsed_and_order(tmp_path):
     runs_db.start_job("young", "b", started_at=_ago(0.01), main_context=ctx)
     runs_db.start_job("old", "a", started_at=_ago(1), main_context=ctx)
     jobs = runs_db.active_jobs(ctx)
-    assert [j["session_id"] for j in jobs] == ["young", "old"]  # youngest first
+    assert [j["session_id"]
+            for j in jobs] == ["young", "old"]  # youngest first
     assert jobs[1]["elapsed"] >= 3500  # ~1h old
 
 
@@ -262,7 +290,8 @@ def test_set_job_session_repoints_a_live_job(tmp_path):
 
     runs_db.set_job_session(job_id, "codex-thread", main_context=ctx)
 
-    assert runs_db.active_jobs(main_context=ctx)[0]["session_id"] == "codex-thread"
+    assert runs_db.active_jobs(main_context=ctx)[
+        0]["session_id"] == "codex-thread"
 
 
 def test_set_job_session_is_a_no_op_without_a_job(tmp_path):
@@ -280,7 +309,8 @@ def test_a_noted_agent_session_is_what_the_run_records(tmp_path):
     runs_db.record_run("skill-a", "issue", "codee-sid", "succeeded",
                        main_context=ctx)
 
-    assert runs_db.recent_runs(main_context=ctx)[0]["session_id"] == "codex-thread"
+    assert runs_db.recent_runs(main_context=ctx)[
+        0]["session_id"] == "codex-thread"
 
 
 def test_the_note_is_consumed_so_a_later_run_keeps_its_own_id(tmp_path):

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee/test_admin_service.py

@@ -182,23 +182,27 @@ class TestAgentConversationTest(unittest.TestCase):
         agent.run.side_effect = run
 
         response, session_id = self.service.test_agent_conversation(
-            "codex", "Hello")
+            "codex", "Hello", model="gpt-6-astra")
 
         self.assertEqual(response, "Final response")
         self.assertEqual(session_id, "agent-thread")
         build_agent.assert_called_once_with(
             self.service.context.settings, Path("/repo"), CodingAgent.CODEX)
+        self.assertEqual(agent.run.call_args.args[2], "gpt-6-astra")
 
     @patch("codee.admin_service.build_coding_agent")
     def test_later_turn_resumes_the_same_session(self, build_agent) -> None:
         build_agent.return_value.continue_conversation.return_value = "Still here"
 
         response, session_id = self.service.test_agent_conversation(
-            "github_copilot", "What did I ask?", "same-thread")
+            "github_copilot", "What did I ask?", "same-thread",
+            "claude-opus-5")
 
         self.assertEqual((response, session_id), ("Still here", "same-thread"))
         call = build_agent.return_value.continue_conversation.call_args
-        self.assertEqual(call.args[:2], ("What did I ask?", "same-thread"))
+        self.assertEqual(
+            call.args[:3],
+            ("What did I ask?", "same-thread", "claude-opus-5"))
 
 
 class AdminServiceWorkItemsTest(unittest.TestCase):

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee_agent_abstract/provider.py

@@ -19,6 +19,17 @@ class AgentModel:
     name: str
 
 
+class AgentResponse(str):
+    """Agent text with optional diagnostic output for the run record."""
+
+    debug_logs: str
+
+    def __new__(cls, response: str, debug_logs: str = "") -> "AgentResponse":
+        value = super().__new__(cls, response)
+        value.debug_logs = debug_logs
+        return value
+
+
 class AbstractCodingAgent(ABC):
     """Base class every coding agent (e.g. Claude Code) inherits from.
 

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee_agent_github_copilot/provider.py

@@ -7,7 +7,7 @@ import time
 from collections.abc import Callable
 from pathlib import Path
 
-from codee_agent_abstract.provider import AbstractCodingAgent, AgentModel
+from codee_agent_abstract.provider import AbstractCodingAgent, AgentModel, AgentResponse
 from codee_main_context.context import Settings
 from codee_main_context.logging import get_logger
 
@@ -30,6 +30,7 @@ _SESSION_NEW_ID = 2
 # Ceiling on what one run may spend. AI credits bill at $0.04 each, so the
 # default is the same $20 cap the Claude Code agent puts on a run.
 MAX_AI_CREDITS_ENV_VAR = "CODEE_COPILOT_MAX_AI_CREDITS"
+COPILOT_DEBUG_ENV_VAR = "COPILOT_DEBUG"
 DEFAULT_MAX_AI_CREDITS = 1000
 # The CLI rejects anything lower outright, which reads as a broken agent rather
 # than a misconfigured cap, so a smaller override is raised to it instead.
@@ -119,6 +120,10 @@ class GitHubCopilotAgent(AbstractCodingAgent):
         # body alone — so a skill's model only takes effect via this flag.
         if model:
             cmd += ["--model", model]
+        debug_enabled = os.environ.get(
+            COPILOT_DEBUG_ENV_VAR, "").strip().lower() == "true"
+        if debug_enabled:
+            cmd += ["--log-level", "debug"]
 
         log.info("Running copilot with message: %s", user_message)
         log.debug("cwd=%s cmd=%s", self._cwd, " ".join(cmd))
@@ -156,7 +161,7 @@ class GitHubCopilotAgent(AbstractCodingAgent):
             # rather than failing a run that the CLI itself called successful.
             log.warning(
                 "copilot produced no result event; returning raw output")
-            return stdout
+            return AgentResponse(stdout, stderr if debug_enabled else "")
         if outcome.get("exitCode"):
             raise RuntimeError(
                 f"Copilot run errored (exit code {outcome['exitCode']}): "
@@ -166,7 +171,7 @@ class GitHubCopilotAgent(AbstractCodingAgent):
             raise RuntimeError(
                 f"Copilot run produced no response: {_detail(errors, stderr)}"
             )
-        return reply
+        return AgentResponse(reply, stderr if debug_enabled else "")
 
 
 def _relative(path: Path, cwd: Path) -> str:

{codee_agent-0.6.5 → codee_agent-0.6.7}/src/codee_agent_github_copilot/test.py

@@ -7,7 +7,8 @@ from pathlib import Path
 from unittest.mock import Mock, patch
 
 from codee_agent_github_copilot.provider import (
-    MAX_AI_CREDITS_ENV_VAR, GitHubCopilotAgent, _await_result)
+    COPILOT_DEBUG_ENV_VAR, MAX_AI_CREDITS_ENV_VAR, GitHubCopilotAgent,
+    _await_result)
 from codee_main_context.context import Settings
 
 SESSION = "82232f47-df60-4cb3-8c3a-de12074c9205"
@@ -153,6 +154,26 @@ class CopilotRunTest(unittest.TestCase):
 
         self.assertNotIn("--model", self.captured.call_args.args[0])
 
+    def test_debug_env_adds_log_level_and_preserves_stderr(self) -> None:
+        stdout = _stream(_event("assistant.message", {"content": "ok"}),
+                         _event("result", exitCode=0))
+        with patch.dict(os.environ, {COPILOT_DEBUG_ENV_VAR: "true"}):
+            response = self._run(_completed(stdout, stderr="debug line\n"))
+
+        cmd = self.captured.call_args.args[0]
+        self.assertEqual(cmd[cmd.index("--log-level") + 1], "debug")
+        self.assertEqual(response, "ok")
+        self.assertEqual(response.debug_logs, "debug line\n")
+
+    def test_debug_env_must_be_true(self) -> None:
+        stdout = _stream(_event("assistant.message", {"content": "ok"}),
+                         _event("result", exitCode=0))
+        with patch.dict(os.environ, {COPILOT_DEBUG_ENV_VAR: "1"}):
+            response = self._run(_completed(stdout, stderr="not retained"))
+
+        self.assertNotIn("--log-level", self.captured.call_args.args[0])
+        self.assertEqual(response.debug_logs, "")
+
     def test_the_best_model_is_an_anthropic_catalog_id(self) -> None:
         # Copilot has no latest-tier alias, so this id is pinned by hand and only
         # a real catalog id will be accepted by the CLI.
@@ -188,6 +209,13 @@ class CopilotRunTest(unittest.TestCase):
         self.assertEqual(self._run(_completed("plain text reply\n")),
                          "plain text reply\n")
 
+    def test_raw_output_keeps_debug_logs(self) -> None:
+        with patch.dict(os.environ, {COPILOT_DEBUG_ENV_VAR: "TRUE"}):
+            response = self._run(_completed("plain text reply\n", "debug\n"))
+
+        self.assertEqual(response, "plain text reply\n")
+        self.assertEqual(response.debug_logs, "debug\n")
+
     def test_a_timeout_raises(self) -> None:
         with patch("subprocess.run",
                    side_effect=subprocess.TimeoutExpired(cmd="copilot", timeout=7200)):