Merge branch 'main' into codex/issue-36-timeout-image

Author: alphacrack

CHANGELOG.md

@@ -5,6 +5,30 @@ All notable changes to this project are documented here.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.6.0] — 2026-07-12
+
+### Added
+- `--dry-run` flag on `run`: stop right after the ingest/planning stage with
+  the feasibility verdict and blockers, before any agent time is spent; the
+  remaining stages are marked skipped with reason "dry-run stop" (#29, #30,
+  thanks @Garifallos). Promoted to a first-class `dry_run` config field, so
+  it can also be set in `readme2demo.toml`.
+
+### Fixed
+- `report --json` crashed with an AttributeError on any run whose manifest
+  had recorded stages (the stages dict was iterated as a list of objects),
+  and reported `cost` as 0.0 and `commit` as null from nonexistent field
+  names — a regression that slipped in alongside the dry-run CLI change
+  (the old test only covered an empty manifest). Restored the real manifest
+  fields (`total_cost_usd`, `commit_sha`) with a regression test that uses a
+  populated manifest.
+
+### Changed
+- CI: bump `actions/checkout` 4 → 7 and `actions/deploy-pages` 4 → 5
+  (Dependabot).
+- Internal go-to-market notes (`launch/`, `RELEASE_PLAN.md`,
+  `IMPLEMENTATION_PLAN.md`) removed from the tracked repo.
+
 ## [0.5.0] — 2026-07-10
 
 ### Added

CITATION.cff

@@ -14,8 +14,8 @@ authors:
 repository-code: "https://github.com/alphacrack/readme2demo"
 url: "https://github.com/alphacrack/readme2demo"
 license: MIT
-version: 0.4.1
-date-released: "2026-07-09"
+version: 0.6.0
+date-released: "2026-07-12"
 keywords:
   - ai-agents
   - developer-tools

docs/usage.md

@@ -34,6 +34,7 @@ provide both. At least one is required.
 | `--gemini [model]` | Run the whole session on Google Gemini via `GEMINI_API_KEY` (OpenHands agent + Gemini passes). Bare flag reads `--model` / `GEMINI_MODEL`. |
 | `--anthropic [model]` | Run the sandboxed agent on OpenHands with a Claude model via `ANTHROPIC_API_KEY`. |
 | `--budget-usd` | Abort the run if the agent's cost exceeds this. |
+| `--dry-run` | Stop after ingest/planning with the feasibility verdict and blockers — no agent time spent. |
 | `--skip-video` / `--with-video` | Skip or force VHS rendering. |
 | `--allow-docker-socket` | Mount the host Docker socket into the sandbox. Security tradeoff — trusted repos only. |
 

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "readme2demo"
-version = "0.5.0"
+version = "0.6.0"
 description = "Verified tutorials and demo videos from your README. An AI agent runs it in a hardened Docker sandbox and replays it in a fresh container before anything is published."
 readme = "README.md"
 requires-python = ">=3.10"

readme2demo.toml.example

@@ -28,6 +28,7 @@ pids_limit = 512
 allow_docker_socket = false
 
 # --- Stages -----------------------------------------------------------------
+dry_run = false               # stop after ingest/planning (feasibility verdict only)
 verify_timeout_s = 900
 verify_retries = 1
 distill_retries = 1

src/readme2demo/cli.py

@@ -335,10 +335,10 @@ def run(
              "else the config default.",
     ),
     config_file: Optional[Path] = typer.Option(None, "--config", help="readme2demo.toml path"),
-    # 1. ΠΡΟΣΘΗΚΗ ΕΔΩ: Δηλώνουμε το option στο Typer
     dry_run: bool = typer.Option(
         False, "--dry-run",
-        help="Run ingest/plan only and print feasibility/blockers, then stop.",
+        help="Run ingest/plan only and print feasibility/blockers, then stop "
+             "— a cheap feasibility check before spending agent time.",
     ),
 ) -> None:
     """Run the full pipeline against a repository, a step-by-step guide, or both."""
@@ -354,11 +354,8 @@ def run(
         config_file, engine, model, output_dir, timeout,
         budget_usd, max_turns, skip_video, base_image, llm_backend,
     )
-    
-    # 2. ΕΔΩ (όπως το είχες): Ενημερώνουμε το config
     if dry_run:
         cfg = cfg.model_copy(update={"dry_run": True})
-        
     if step_by_step is not None:
         cfg = cfg.model_copy(update={"step_by_step": step_by_step})
     if repo_url is None:
@@ -448,16 +445,19 @@ def resume(
 @app.command()
 def report(
     run_dir: Path = typer.Argument(..., help="Path to a runs/<run-id> directory"),
-   json_output: bool = typer.Option(False, "--json", help="Emit summary as JSON"),
+    json_output: bool = typer.Option(False, "--json", help="Emit summary as JSON"),
 ) -> None:
     """Print a summary of a run: stage statuses, verification, cost."""
     manifest = Manifest.load(run_dir)
     if json_output:
         output_data = {
-            "stages": [{"name": s.name, "status": s.status} for s in getattr(manifest, "stages", [])] if hasattr(manifest, "stages") else {},
-            "verified": getattr(manifest, "verified", False),
-            "cost": getattr(manifest, "cost", 0.0),
-            "commit": getattr(manifest, "commit", None) or getattr(manifest, "repo_commit", None),
+            "stages": [
+                {"name": name, "status": rec.status}
+                for name, rec in manifest.stages.items()
+            ],
+            "verified": manifest.verified,
+            "cost": manifest.total_cost_usd,
+            "commit": manifest.commit_sha,
         }
         print(json.dumps(output_data, indent=2))
         raise typer.Exit(0)

src/readme2demo/config.py

@@ -39,6 +39,9 @@ class Config(BaseModel):
     pids_limit: int = 512
 
     # Stages
+    # --dry-run: stop after ingest/planning (feasibility verdict + blockers),
+    # skipping the paid agent stage and everything downstream.
+    dry_run: bool = False
     verify_timeout_s: int = 900
     verify_retries: int = 1  # plain script retries before distiller feedback loop
     distill_retries: int = 1  # distiller feedback loops on verify failure

src/readme2demo/orchestrator.py

@@ -333,10 +333,13 @@ def run(self) -> Manifest:
                 self.manifest.stage_fail(stage, f"{type(e).__name__}: {e}")
                 raise
 
-            # Έλεγχος για dry-run αμέσως μετά την επιτυχrunning του ingest στάδιου
-            if stage == "ingest" and getattr(self.cfg, "dry_run", False):
-                console.print("\n[bold yellow]ℹ Dry-run mode active. Stopping after ingest and planning.[/]")
-                # Μαρκάρουμε τα επόμενα στάδια ως skipped με σαφή αιτιολογία
+            # --dry-run: the feasibility verdict and blockers are known once
+            # ingest completes; stop before any agent cost is spent and mark
+            # the remaining stages skipped with a clear reason.
+            if stage == "ingest" and self.cfg.dry_run:
+                console.print(
+                    "\n[bold yellow]ℹ Dry run: stopping after ingest/planning.[/]"
+                )
                 for s in ("agent", "normalize", "distill", "verify", "render", "tutorial"):
                     if self.manifest.stages[s].status != "completed":
                         self.manifest.stage_skip(s, reason="dry-run stop")

tests/test_cli.py

@@ -386,27 +386,35 @@ def test_regression_report_keeps_bracketed_error_text(tmp_path):
     assert result.exit_code == 0
     assert "readme2demo[openai]" in result.output
 
-def test_report_json_output(tmp_path, monkeypatch):
+def test_regression_report_json_with_recorded_stages(tmp_path):
+    """Regression: `report --json` crashed with AttributeError on any manifest
+    that had recorded stages (#29 iterated the stages dict as a list of
+    objects), and read nonexistent `cost`/`commit` fields so cost was always
+    0.0 and commit always null. Must emit one entry per stage plus the real
+    total_cost_usd / commit_sha values. (The old test used empty stages, so
+    CI stayed green through the breakage.)
+    """
     import json
 
-  
     manifest_data = {
         "run_id": "test-run-123",
         "verified": True,
-        "cost": 1.50,
-        "repo_commit": "abcdef123456",
-        "stages": {}
+        "total_cost_usd": 1.5,
+        "commit_sha": "abcdef123456",
+        "stages": {
+            "ingest": {"status": "completed"},
+            "agent": {"status": "failed"},
+        },
     }
-    manifest_file = tmp_path / "manifest.json"
-    manifest_file.write_text(json.dumps(manifest_data))
+    (tmp_path / "manifest.json").write_text(json.dumps(manifest_data))
 
     result = runner.invoke(app, ["report", str(tmp_path), "--json"])
-    
+
     assert result.exit_code == 0
-    
-    parsed_output = json.loads(result.output)
-    
-    assert parsed_output["verified"] is True
-    assert parsed_output["cost"] == 0.0
-    assert "commit" in parsed_output
+    parsed = json.loads(result.output)
+    assert parsed["verified"] is True
+    assert parsed["cost"] == 1.5
+    assert parsed["commit"] == "abcdef123456"
+    assert {"name": "ingest", "status": "completed"} in parsed["stages"]
+    assert {"name": "agent", "status": "failed"} in parsed["stages"]
 

tests/test_orchestrator.py

@@ -167,3 +167,23 @@ def test_reset_from_after_verify_keeps_verdict(tmp_path: Path):
     assert m.verified is True
     m.reset_from("verify")
     assert m.verified is False
+
+
+def test_dry_run_stops_after_ingest(tmp_path: Path, monkeypatch):
+    # --dry-run: feasibility verdict lands in the manifest, every later stage
+    # is skipped with a clear reason, and no agent stage ever starts.
+    cfg = Config(runs_dir=tmp_path, dry_run=True)
+    orch = Orchestrator.new_run("https://github.com/x/y", cfg)
+
+    def fake_ingest(repo_url, run_dir, model, **kwargs):
+        plan = make_plan(feasible=True)
+        (run_dir / "plan.json").write_text(plan.model_dump_json())
+        return plan, "abc1234", 0.005
+
+    monkeypatch.setattr(ingest_mod, "ingest", fake_ingest)
+    manifest = orch.run()
+    assert manifest.stages["ingest"].status == "completed"
+    for s in ("agent", "normalize", "distill", "verify", "render", "tutorial"):
+        assert manifest.stages[s].status == "skipped"
+        assert manifest.stages[s].meta.get("reason") == "dry-run stop"
+    assert manifest.verified is False