refactorai-cli 0.3.0__tar.gz → 0.3.2__tar.gz

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: refactorai-cli
-Version: 0.3.0
+Version: 0.3.2
 Summary: Local-first CLI for the refactor platform
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown
@@ -8,7 +8,7 @@ Requires-Dist: typer>=0.12.0
 Requires-Dist: httpx>=0.27.0
 Requires-Dist: rich>=13.7.0
 Requires-Dist: PyYAML>=6.0.1
-Requires-Dist: refactorai-core>=3.0.0
+Requires-Dist: refactorai-core>=3.0.2
 
 # refactorai-cli
 
@@ -54,7 +54,7 @@ python -m twine upload ./refactorai-cli/dist/*
 ## Install test (local)
 
 ```bash
-python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.0-py3-none-any.whl
+python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.2-py3-none-any.whl
 refactor --version
 ```
 

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/README.md

@@ -42,7 +42,7 @@ python -m twine upload ./refactorai-cli/dist/*
 ## Install test (local)
 
 ```bash
-python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.0-py3-none-any.whl
+python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.2-py3-none-any.whl
 refactor --version
 ```
 

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "refactorai-cli"
-version = "0.3.0"
+version = "0.3.2"
 description = "Local-first CLI for the refactor platform"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -12,7 +12,7 @@ dependencies = [
   "httpx>=0.27.0",
   "rich>=13.7.0",
   "PyYAML>=6.0.1",
-  "refactorai-core>=3.0.0",
+  "refactorai-core>=3.0.2",
 ]
 
 [project.scripts]

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/refactorai_cli/__init__.py

@@ -5,4 +5,4 @@ the shared `refactor_core` pipeline from a project folder while staying
 authenticated to the hosted platform via a developer key.
 """
 
-__version__ = "0.3.0"
+__version__ = "0.3.2"

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/refactorai_cli/commands/run_cmds.py

@@ -157,19 +157,25 @@ def _default_config_for_init() -> tuple[str, str]:
     if backend == BACKEND_LOCAL_SERVER:
         model_id = _setup_recommended_model_id()
         base_url = _setup_engine_base_url()
-        config_text = DEFAULT_CONFIG.replace("provider: heuristic", "provider: ollama", 1)
-        config_text = config_text.replace(
+        # local_server stays in `mode: local` but swaps the default heuristic
+        # provider block for an Ollama block pointing at the local engine.
+        config_text = DEFAULT_CONFIG.replace(
+            "# Heuristic (local, no LLM) - default for mode: local\n"
+            "provider: heuristic\n"
             "model_id: heuristic-local-v1",
-            f"model_id: {model_id}\nbase_url: {base_url}",
+            "# Ollama (local engine) - configured by `refactor init`\n"
+            "provider: ollama\n"
+            f"model_id: {model_id}\n"
+            f"base_url: {base_url}",
             1,
         )
         return config_text, "local_server"
     if backend == BACKEND_BYOK:
-        config_text = DEFAULT_CONFIG.replace("mode: local", "mode: cloud_byok", 1)
-        config_text = config_text.replace("provider: heuristic", "provider: openai", 1)
-        config_text = config_text.replace(
-            "model_id: heuristic-local-v1",
-            "model_id: gpt-4.1-mini\nprovider_key: ${OPENAI_API_KEY}\n# credential_ref: cred_01J...",
+        # BYOK only flips the execution mode. The developer chooses and
+        # uncomments a provider block in the Provider selection section.
+        config_text = DEFAULT_CONFIG.replace(
+            "mode: local\n# mode: cloud_byok",
+            "# mode: local\nmode: cloud_byok",
             1,
         )
         return config_text, "byok"
@@ -659,29 +665,7 @@ def _cloud_inference(
     }
 
     if execution_mode == "cloud_byok":
-        provider_name = str(constitution.get_setting("provider", "") or "").strip()
-        provider_key = str(constitution.get_setting("provider_key", "") or "").strip()
-        credential_ref = str(constitution.get_setting("credential_ref", "") or "").strip()
-
-        # Env-key-first BYOK path: when provider_key is configured (typically via
-        # ${ENV_VAR} interpolation), prefer it over credential_ref.
-        if provider_key:
-            if not provider_name or provider_name.lower() == "cloud":
-                raise RuntimeError(
-                    "cloud_byok with provider_key requires an explicit provider "
-                    "(e.g. provider: openai)."
-                )
-            payload["provider_key"] = provider_key
-            payload["provider"] = provider_name
-        elif credential_ref:
-            payload["credential_ref"] = credential_ref
-            if provider_name and provider_name.lower() != "cloud":
-                payload["provider"] = provider_name
-        else:
-            raise RuntimeError(
-                "cloud_byok requires either `provider_key` (preferred env-key path) "
-                "or `credential_ref` (platform-stored credential path) in refactor.config."
-            )
+        _apply_byok_payload(payload, constitution)
 
     idempotency_key = _stable_idempotency_key(
         account_id=auth_ctx.account_id,
@@ -758,6 +742,185 @@ def _cloud_inference(
     )
 
 
+def _apply_byok_payload(payload: dict, constitution) -> None:
+    """Populate BYOK provider fields on a cloud inference ``payload`` in place.
+
+    Env-key-first: prefer ``provider_key`` (typically ``${ENV_VAR}``) over a
+    platform-stored ``credential_ref``. Shared by the review/refactor path and
+    the per-RR proposal path so BYOK resolution stays consistent.
+    """
+    provider_name = str(constitution.get_setting("provider", "") or "").strip()
+    provider_key = str(constitution.get_setting("provider_key", "") or "").strip()
+    credential_ref = str(constitution.get_setting("credential_ref", "") or "").strip()
+
+    if provider_key:
+        if not provider_name or provider_name.lower() == "cloud":
+            raise RuntimeError(
+                "cloud_byok with provider_key requires an explicit provider "
+                "(e.g. provider: openai)."
+            )
+        payload["provider_key"] = provider_key
+        payload["provider"] = provider_name
+    elif credential_ref:
+        payload["credential_ref"] = credential_ref
+        if provider_name and provider_name.lower() != "cloud":
+            payload["provider"] = provider_name
+    else:
+        raise RuntimeError(
+            "cloud_byok requires either `provider_key` (preferred env-key path) "
+            "or `credential_ref` (platform-stored credential path) in refactor.config."
+        )
+
+
+def _post_cloud_inference(*, payload: dict, idempotency_key: str, auth_ctx, execution_mode: str) -> dict:
+    """POST a cloud inference request and return the parsed JSON body.
+
+    Centralizes transport + status-code error mapping so the per-RR proposal
+    path surfaces the same remediation messages as review/refactor.
+    """
+    try:
+        response = httpx.post(
+            f"{platform_url()}/v1/inference/refactor",
+            headers={
+                "Authorization": f"Bearer {auth_ctx.key.key}",
+                "Content-Type": "application/json",
+                "Idempotency-Key": idempotency_key,
+            },
+            json=payload,
+            timeout=45.0,
+        )
+    except httpx.HTTPError as exc:
+        raise RuntimeError(f"Could not reach cloud inference service: {exc}") from exc
+
+    if response.status_code == 401:
+        raise RuntimeError("Cloud inference unauthorized (developer key invalid or revoked).")
+    if response.status_code == 402:
+        if execution_mode == "cloud_byok":
+            raise RuntimeError(
+                "Cloud BYOK is not enabled for this account (missing `run:byok`). "
+                "Contact your admin to enable BYOK cloud execution."
+            )
+        raise RuntimeError("Cloud inference requires paid tier (`mode: cloud_managed`).")
+    if response.status_code == 404 and execution_mode == "cloud_byok":
+        raise RuntimeError(
+            "BYOK credential not found or not owned by this account. "
+            "Check `refactor cloud credentials list` and your `credential_ref`."
+        )
+    if response.status_code == 429:
+        raise RuntimeError("Cloud inference quota exhausted or rate-limited.")
+    if response.status_code >= 400:
+        raise RuntimeError(f"Cloud inference failed ({response.status_code}): {response.text[:500]}")
+
+    return response.json() if response.content else {}
+
+
+class _CloudRRProvider:
+    """Lightweight provider stub for cloud RR execution (R22).
+
+    The executor reads ``provider.model_id`` for the artifact and would call
+    ``provider.complete(...)`` for proposals/tests. In cloud mode the proposal
+    and test-gen hooks are supplied, so ``complete`` is never invoked; it raises
+    to make any accidental local-prompt path explicit rather than silent.
+    """
+
+    def __init__(self, *, model_id: str) -> None:
+        self.model_id = model_id
+        self.provider_name = "cloud"
+        self.supports_refactor = True
+
+    def complete(self, system: str, user: str) -> str:  # pragma: no cover - guard
+        raise RuntimeError(
+            "cloud RR execution must use server-side proposal hooks; "
+            "provider.complete should not be called."
+        )
+
+
+def _cloud_rr_hooks(
+    *,
+    project_root: Path,
+    constitution,
+    model_hint: str | None,
+) -> tuple[Callable, Callable, str]:
+    """Build (propose_fn, gen_test_fn, model_id) for cloud per-RR execution (R22).
+
+    The returned callables match the executor's hook contract. They send the
+    single RR's context + current source to the platform, which builds the
+    proprietary prompts server-side and returns a proposal / characterization
+    test. Idempotency keys include the RR id and attempt so retries of distinct
+    RRs/attempts never collapse into a replayed result.
+    """
+    execution_mode = _cloud_execution_mode(constitution) or "cloud_managed"
+    auth_ctx = ensure_authenticated(project_root)
+    constitution_body = getattr(constitution, "body", "") or ""
+    resolved_model = model_hint or str(
+        constitution.get_setting("model_id", "managed-default") or "managed-default"
+    )
+
+    def _rr_context(rr) -> dict:
+        return {
+            "location": getattr(rr, "location", ""),
+            "file_path": getattr(rr, "file_path", ""),
+            "category": getattr(rr, "category", ""),
+            "severity": getattr(rr, "severity", ""),
+            "recommendation": getattr(rr, "recommendation", ""),
+        }
+
+    def _base_payload(stage: str, rr, source: str) -> dict:
+        payload = {
+            "stage": stage,
+            "execution_mode": execution_mode,
+            "constitution_hash": constitution.content_hash,
+            "constitution_body": constitution_body,
+            "rr_context": _rr_context(rr),
+            "source": source,
+            "max_alerts": int(constitution.get_setting("max_alerts_per_run", 5) or 5),
+            "model_hint": resolved_model,
+        }
+        if execution_mode == "cloud_byok":
+            _apply_byok_payload(payload, constitution)
+        return payload
+
+    def _idempotency(stage: str, rr, attempt: int) -> str:
+        seed = "|".join(
+            [
+                auth_ctx.account_id,
+                execution_mode,
+                stage,
+                constitution.content_hash,
+                str(getattr(rr, "rr_id", "")),
+                str(int(attempt)),
+            ]
+        )
+        return hashlib.sha256(seed.encode("utf-8")).hexdigest()
+
+    def propose_fn(rr, source: str, prior_failures: list[str]) -> dict:
+        # Attempt number is derived from prior failures so each retry gets a
+        # distinct idempotency key (1-based).
+        attempt = len(prior_failures or []) + 1
+        payload = _base_payload("rr_propose", rr, source)
+        payload["prior_failures"] = list(prior_failures or [])
+        data = _post_cloud_inference(
+            payload=payload,
+            idempotency_key=_idempotency("rr_propose", rr, attempt),
+            auth_ctx=auth_ctx,
+            execution_mode=execution_mode,
+        )
+        proposal = data.get("proposal") or {}
+        if not isinstance(proposal, dict):
+            return {}
+        return proposal
+
+    def gen_test_fn(rr, source: str) -> dict:
+        # Server-side characterization-test generation is not part of Phase 1.
+        # Returning no test makes the executor fall back to the configured test
+        # command for behavioral verification (parity with a local run that has
+        # no generated test). This keeps cloud RR execution safe today and lets
+        # a future phase add a dedicated server test-gen stage.
+        return {}
+
+    return propose_fn, gen_test_fn, resolved_model
+
+
 def _load_project() -> tuple[Path, object, object]:
     files = find_project_files()
     if not files:
@@ -1796,7 +1959,8 @@ def review(
         if _sandbox_enabled(project_config):
             if cloud:
                 console.print(
-                    "[dim]review: sandbox.enabled is ignored in cloud_managed mode.[/dim]"
+                    f"[dim]review: sandbox.enabled is ignored in "
+                    f"{_cloud_execution_mode(constitution)} mode.[/dim]"
                 )
             else:
                 sandbox_ctx = _ensure_sandbox_ready(
@@ -2023,7 +2187,8 @@ def code(
         if _sandbox_enabled(project_config):
             if cloud:
                 console.print(
-                    "[dim]code: sandbox.enabled is ignored in cloud_managed mode.[/dim]"
+                    f"[dim]code: sandbox.enabled is ignored in "
+                    f"{_cloud_execution_mode(constitution)} mode.[/dim]"
                 )
             else:
                 sandbox_ctx = _ensure_sandbox_ready(
@@ -2127,64 +2292,23 @@ def code(
             console.print("[green]No issues to refactor.[/green]")
             return
 
-        if cloud:
-            # Cloud modes generate patches via the platform inference API rather
-            # than the local RR interactive loop. Reuse prior review findings as
-            # prompt context; fall back to the latest stored review findings when
-            # requests were carried over from an earlier run.
-            if not cloud_review_findings:
-                existing_dir = head_run_dir(project_root)
-                if existing_dir is not None:
-                    try:
-                        cloud_review_findings = [
-                            Finding.model_validate(item) for item in (read_findings(existing_dir) or [])
-                        ]
-                    except Exception:
-                        cloud_review_findings = []
-            with console.status("[cyan]code[/cyan] generating patches (cloud)...", spinner="dots"):
-                artifact = _cloud_inference(
-                    stage="refactor",
-                    project_root=project_root,
-                    constitution=constitution,
-                    target=target,
-                    model_hint=provider,
-                    findings=cloud_review_findings,
-                )
-            cloud_store = create_run(project_root, run_id=artifact.run_id)
-            cloud_store.write_artifact(artifact, target=target, status="completed")
+        from refactor_core.execution.rr_executor import run_refactor_requests
 
-            _print_rr_summary(artifact)
-            _print_patches(artifact.patches)
-            gate_color = "green" if artifact.gate.status == "passed" else "red"
-            console.print(
-                f"\n[bold]result[/bold]: [{gate_color}]{artifact.gate.status}[/{gate_color}]  "
-                f"[dim]model={artifact.model_id} mode={_cloud_execution_mode(constitution)} "
-                f"run={artifact.run_id}[/dim]"
+        # Cloud and local both drive the SAME per-RR interactive loop (R22 cloud
+        # RR parity). The only difference is where proposals come from: cloud
+        # passes server-side proposal hooks; local uses the local provider and
+        # client-side prompt builders. This guarantees identical UX (proposal
+        # diff + approval prompt), status transitions, and outcomes table.
+        cloud_propose_fn = None
+        cloud_gen_test_fn = None
+        rr_provider = selected_provider
+        if cloud:
+            cloud_propose_fn, cloud_gen_test_fn, cloud_model_id = _cloud_rr_hooks(
+                project_root=project_root,
+                constitution=constitution,
+                model_hint=provider,
             )
-
-            if apply:
-                gate_mode = str(constitution.get_setting("gate_mode", "strict") or "strict")
-                diffs = [p.diff_text for p in (artifact.patches or [])]
-                if not diffs:
-                    console.print("[dim]No patches to apply.[/dim]")
-                else:
-                    _apply_diffs(
-                        project_root,
-                        cloud_store.run_dir,
-                        diffs,
-                        artifact.gate.status,
-                        gate_mode,
-                        force,
-                        project_config,
-                    )
-            elif artifact.patches:
-                console.print(
-                    "[dim]Review patches with `refactor diff`, then apply with "
-                    "`refactor code --apply` or `refactor apply`.[/dim]"
-                )
-            return
-
-        from refactor_core.execution.rr_executor import run_refactor_requests
+            rr_provider = _CloudRRProvider(model_id=cloud_model_id)
 
         with console.status("[cyan]code[/cyan] running...", spinner="dots") as status:
             rr_config = (
@@ -2195,13 +2319,15 @@ def code(
             artifact = run_refactor_requests(
                 project_root=project_root,
                 constitution=constitution,
-                provider=selected_provider,
+                provider=rr_provider,
                 requests=requests,
                 config=rr_config,
                 progress_cb=_rr_progress_logger(status=status),
                 approval_cb=_rr_approval_prompt(auto_approve=auto_approve),
                 on_rr_failure=on_rr_failure,
                 equivalence_mode=equivalence,
+                propose_fn=cloud_propose_fn,
+                gen_test_fn=cloud_gen_test_fn,
             )
     except typer.Exit:
         # Intentional exits (e.g. gate-blocked apply) keep their own code/message.

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/refactorai_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: refactorai-cli
-Version: 0.3.0
+Version: 0.3.2
 Summary: Local-first CLI for the refactor platform
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown
@@ -8,7 +8,7 @@ Requires-Dist: typer>=0.12.0
 Requires-Dist: httpx>=0.27.0
 Requires-Dist: rich>=13.7.0
 Requires-Dist: PyYAML>=6.0.1
-Requires-Dist: refactorai-core>=3.0.0
+Requires-Dist: refactorai-core>=3.0.2
 
 # refactorai-cli
 
@@ -54,7 +54,7 @@ python -m twine upload ./refactorai-cli/dist/*
 ## Install test (local)
 
 ```bash
-python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.0-py3-none-any.whl
+python -m pip install ./refactorai-cli/dist/refactorai_cli-0.3.2-py3-none-any.whl
 refactor --version
 ```
 

{refactorai_cli-0.3.0 → refactorai_cli-0.3.2}/refactorai_cli.egg-info/requires.txt

@@ -2,4 +2,4 @@ typer>=0.12.0
 httpx>=0.27.0
 rich>=13.7.0
 PyYAML>=6.0.1
-refactorai-core>=3.0.0
+refactorai-core>=3.0.2