duet-cli 0.2.20__tar.gz → 0.2.21__tar.gz

{duet_cli-0.2.20/duet_cli.egg-info → duet_cli-0.2.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: duet-cli
-Version: 0.2.20
+Version: 0.2.21
 Summary: Two CLI agents in conversation. One Python file. Stdlib only.
 Author: Volkan Altan
 License-Expression: MIT
@@ -33,9 +33,9 @@ Codex. Gemini and Copilot are also supported. A native ZCode account can
 provide a [read-only review peer](https://github.com/volkan/duet/blob/main/docs/ZCODE.md).
 [See examples using one CLI](https://github.com/volkan/duet/blob/main/docs/USAGE.md#same-backend-peering).
 
-[![Watch the 80-second review workflow demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://github.com/volkan/duet/blob/main/docs/demos/finding-review.mp4)
+[![Watch the review demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://volkan.github.io/duet/demos/)
 
-An edited replay using a public toy fixture. [Playback and source](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
+An edited replay using a public toy fixture. [Demo details and evidence](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
 
 ## A real example
 

{duet_cli-0.2.20 → duet_cli-0.2.21}/README.md

@@ -12,9 +12,9 @@ Codex. Gemini and Copilot are also supported. A native ZCode account can
 provide a [read-only review peer](https://github.com/volkan/duet/blob/main/docs/ZCODE.md).
 [See examples using one CLI](https://github.com/volkan/duet/blob/main/docs/USAGE.md#same-backend-peering).
 
-[![Watch the 80-second review workflow demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://github.com/volkan/duet/blob/main/docs/demos/finding-review.mp4)
+[![Watch the review demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://volkan.github.io/duet/demos/)
 
-An edited replay using a public toy fixture. [Playback and source](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
+An edited replay using a public toy fixture. [Demo details and evidence](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
 
 ## A real example
 

{duet_cli-0.2.20 → duet_cli-0.2.21}/duet.py

@@ -74,7 +74,7 @@ from typing import Callable, Iterable, Optional
 
 # ---------- defaults ----------
 
-__version__ = "0.2.20"
+__version__ = "0.2.21"
 
 DEFAULT_SENTINEL = "<<<LGTM>>>"
 DEFAULT_SANDBOX = "workspace-write"
@@ -6230,7 +6230,7 @@ _RUN_ID_RE = re.compile(r"^\d{8}-\d{6}(?:-\d+)?$")
 
 # ---------- structured finding reports ----------
 
-_FINDING_ID_RE = re.compile(r"\A[LP][1-9][0-9]{0,3}\Z")
+_FINDING_ID_RE = re.compile(r"\A[A-Za-z0-9][A-Za-z0-9._-]{0,31}\Z")
 _FINDING_BLOCK_MAX_BYTES = 64 * 1024
 _FINDING_EVENT_LIMIT = 1000
 _FINDING_DISPOSITIONS = frozenset({"supported", "refuted", "unresolved"})
@@ -6497,7 +6497,8 @@ def _finding_code(value: object) -> str:
 
 
 def _render_finding(finding: dict) -> str:
-    parts = [f"## {finding['id']} — {finding['disposition']} (agent assessment)",
+    # Inline code keeps IDs such as www.example.com or a._b_ from rendering as a link or emphasis.
+    parts = [f"## `{finding['id']}` — {finding['disposition']} (agent assessment)",
              _finding_code(finding["claim"])]
     for event in finding["assessments"]:
         origin = "inherited" if event["inherited"] else "current run"
@@ -7721,8 +7722,14 @@ def _finding_prompt_addendum(agent: Agent, cfg: DuetConfig) -> str:
         '"disposition":"supported|refuted|unresolved","evidence":["file:line or '
         'concrete check and result"],"objection":"remaining concern or null"}. '
         "Use actual enum values, not the pipe-separated alternatives. Preserve IDs "
-        "from the prior conversation; allocate new IDs with your prefix " + prefix +
-        " and an unused positive integer. Reassess disputed findings explicitly. "
+        "from the prior conversation. A valid ID has 1 to 32 ASCII characters: it "
+        "starts with a letter or digit, then uses letters, digits, '.', '_', or '-'. "
+        "When the task, the kickoff, or the other agent already labels a claim with a "
+        "valid ID, reuse that exact ID, including its case. Otherwise allocate a new "
+        "ID with your prefix " + prefix + " and an unused positive integer. Never "
+        "copy a label that is not a valid ID, such as '#1' or 'Finding 3': allocate "
+        "a new ID and name that label in the claim. Reassess disputed findings "
+        "explicitly. "
         "Use unresolved when evidence is missing. Cite only evidence you inspected; "
         "never invent executed checks or human acceptance. An empty findings array "
         "means no finding updates, not proof that the review is correct. Keep recap "

{duet_cli-0.2.20 → duet_cli-0.2.21/duet_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: duet-cli
-Version: 0.2.20
+Version: 0.2.21
 Summary: Two CLI agents in conversation. One Python file. Stdlib only.
 Author: Volkan Altan
 License-Expression: MIT
@@ -33,9 +33,9 @@ Codex. Gemini and Copilot are also supported. A native ZCode account can
 provide a [read-only review peer](https://github.com/volkan/duet/blob/main/docs/ZCODE.md).
 [See examples using one CLI](https://github.com/volkan/duet/blob/main/docs/USAGE.md#same-backend-peering).
 
-[![Watch the 80-second review workflow demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://github.com/volkan/duet/blob/main/docs/demos/finding-review.mp4)
+[![Watch the review demo](https://raw.githubusercontent.com/volkan/duet/main/docs/demos/finding-review.png)](https://volkan.github.io/duet/demos/)
 
-An edited replay using a public toy fixture. [Playback and source](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
+An edited replay using a public toy fixture. [Demo details and evidence](https://github.com/volkan/duet/blob/main/docs/demos/README.md).
 
 ## A real example
 

{duet_cli-0.2.20 → duet_cli-0.2.21}/tests/test_finding_reports.py

@@ -47,7 +47,7 @@ class TestFindingParser(unittest.TestCase):
     def test_duplicate_blocks_unclosed_json_and_duplicate_ids_reject_atomically(self):
         for reply in (block([]) + "\n" + block([]), "```duet-findings\n{}",
                       "```duet-findings\nnot JSON\n```", block([item(), item()]),
-                      block([item(), item("bad")])):
+                      block([item(), item("bad id")])):
             with self.subTest(reply=reply):
                 self.assertEqual(duet.parse_finding_updates(reply), {"status": "malformed", "items": []})
 
@@ -56,8 +56,18 @@ class TestFindingParser(unittest.TestCase):
         self.assertEqual(duet.parse_finding_updates(reply)["status"], "missing")
         self.assertEqual(duet.parse_finding_updates(reply + "\n" + block([]))["status"], "ok")
 
+    def test_caller_defined_ids_are_accepted_and_bad_ids_rejected(self):
+        valid = ["F1", "C1", "P1-1", "a.b_c-d", "9", "L" + "x" * 31]
+        parsed = duet.parse_finding_updates(block([item(identity) for identity in valid]))
+        self.assertEqual([entry["id"] for entry in parsed["items"]], valid)
+        invalid = ["", "F 1", "L" + "x" * 32, "-F1", ".F1", "\u00c9\u00c91", "F1\n",
+                   "_F1", "#1", "Finding 3"]
+        for identity in invalid:
+            with self.subTest(identity=identity):
+                self.assertEqual(duet.parse_finding_updates(block([item(identity)]))["status"], "malformed")
+
     def test_field_limits_and_types(self):
-        invalid = [item("L0"), item("L10000"), item("P01"), item(claim=""),
+        invalid = [item(claim=""),
                    item(claim="x" * 2001), item(disposition=[]), item(evidence="test"),
                    item(evidence=["x"] * 9), item(evidence=["x" * 2001]),
                    item(objection="x" * 2001), item(claim="\ud800")]
@@ -165,6 +175,15 @@ class TestFindingRendering(unittest.TestCase):
         self.assertIn("private", rendered)
         self.assertIn("not proven facts", rendered)
 
+    def test_heading_shows_id_as_inline_code(self):
+        for identity in ["L1", "bug_12", "F1._note_", "www.example.com"]:
+            with self.subTest(identity=identity):
+                report = duet.build_finding_report(state([turn(1, "reviewer", [item(identity)])]))
+                rendered = duet.render_finding_report(report)
+                self.assertIn(f"## `{identity}` \u2014 unresolved (agent assessment)\n", rendered)
+                self.assertEqual(report["findings"][0]["id"], identity)
+                self.assertIn(f"--resolve {identity} --turns 2", rendered)
+
 
 if __name__ == "__main__":
     unittest.main()

{duet_cli-0.2.20 → duet_cli-0.2.21}/tests/test_finding_workflow.py

@@ -137,18 +137,35 @@ class TestFindingWorkflow(unittest.TestCase):
         self.assertNotIn("{SENTINEL}", prompt)
         self.assertIn("duet-findings", prompt)
         self.assertIn('"id":"L1"', prompt)
+        self.assertIn("1 to 32 ASCII characters", prompt)
+        self.assertIn("letters, digits, '.', '_', or '-'", prompt)
+        self.assertIn("starts with a letter or digit", prompt)
+        self.assertIn("reuse that exact ID, including its case", prompt)
+        self.assertIn("Never copy a label that is not a valid ID", prompt)
 
     def test_successful_turns_persist_updates_and_rebuild_local_report(self) -> None:
         state = self._run([_reply("supported"), _reply("supported")])
         run = pathlib.Path(state["transcript_path"]).parent
         saved = json.loads((run / "state.json").read_text(encoding="utf-8"))
         self.assertEqual([entry["finding_updates"]["status"] for entry in saved["history"]], ["ok", "ok"])
-        self.assertIn("L1 — supported", (run / "review.md").read_text(encoding="utf-8"))
+        self.assertIn("`L1` — supported", (run / "review.md").read_text(encoding="utf-8"))
         (run / "review.md").unlink()
         duet._write_run_state(run / "state.json", saved)
         self.assertEqual((run / "review.md").read_text(encoding="utf-8"),
                          duet.render_finding_report(duet.build_finding_report(saved)))
 
+    def test_caller_defined_ids_differing_by_case_stay_separate_through_a_run(self) -> None:
+        def reply(identity: str) -> str:
+            return "```duet-findings\n" + json.dumps(
+                {"findings": [_assessment(identity=identity)]}) + "\n```"
+        state = self._run([reply("F1"), reply("f1")])
+        stored = [entry["finding_updates"]["items"][0]["id"] for entry in state["history"]]
+        self.assertEqual(stored, ["F1", "f1"])
+        run = pathlib.Path(state["transcript_path"]).parent
+        report = (run / "review.md").read_text(encoding="utf-8")
+        self.assertIn("## `F1` — ", report)
+        self.assertIn("## `f1` — ", report)
+
     def test_failed_reply_cannot_publish_structured_claims_from_error_text(self) -> None:
         state = self._run([duet.AgentRunError("agent_error", _reply("supported"))])
         self.assertNotIn("finding_updates", state["history"][0])
@@ -173,7 +190,7 @@ class TestFindingWorkflow(unittest.TestCase):
             self.assertFalse((run / "review.md").exists())
             code, out, err = self._cli("--report", str(run))
             self.assertEqual(code, 0, err)
-            self.assertIn("L1 — supported", out)
+            self.assertIn("`L1` — supported", out)
             self.assertFalse((run / "review.md").exists())
 
     def test_forced_success_and_failure_follow_same_update_contract(self) -> None:
@@ -235,6 +252,16 @@ class TestFindingWorkflow(unittest.TestCase):
         self.assertEqual([finding["id"] for finding in report["findings"]], ["L1"])
         self.assertEqual(report["findings"][0]["disposition"], "refuted")
 
+    def test_resolve_selects_caller_defined_id(self) -> None:
+        history = [{"turn": turn, "agent": agent, "finding_updates": {
+            "status": "ok", "items": [_assessment(identity="P1-1", disposition=disposition)]}}
+            for turn, agent, disposition in ((1, "lead", "supported"), (2, "partner", "unresolved"))]
+        run = self._saved_run("caller-id", history=history, turns_used=2)
+        cfg = self._continue(run, "--resolve", "P1-1")
+        self.assertEqual(cfg.resolve_findings, ["P1-1"])
+        self.assertEqual([event["id"] for event in cfg.finding_baseline], ["P1-1", "P1-1"])
+        self.assertIn("P1-1", cfg.kickoff)
+
     def test_resolve_rejects_unknown_resolved_ids_and_short_budget(self) -> None:
         open_run = self._saved_run("open", history=self._history())
         closed_run = self._saved_run("closed", history=self._history(resolved=True))