Bump version

Author: gltanaka

CHANGELOG.md

@@ -1,35 +1,55 @@
+## v0.0.158 (2026-02-23)
+
+### Feat
+
+- Implement comprehensive API routes for the video pipeline and add a new E2E test for Stage 8 Composition Generation.
+- Add comprehensive Playwright E2E tests for all video editor stages and enhance the script API to support TTS script management.
+- add E2E test for non-recursive circular include detection and update project configuration defaults for video editor.
+- Update test durations and modify `test_llm_invoke_integration.py`.
+
+### Fix
+
+- make sync regression cases 5 and 9 non-fatal on LLM failures
+- handle multi-language sync failure gracefully in case 4
+- increase timeouts and add --max-attempts to sync regression tests
+
+### Refactor
+
+- improve color readability assertion in review tab E2E test by using canvas for RGB extraction.
+- Split audit results API into a dedicated route, implement the Stage 10 Audit UI, and add comprehensive E2E tests including dark theme readability checks.
+- Restructure ProjectConfig schema for output resolution, TTS, audio sync, VEO, and render settings, updating relevant UI and tests.
+- Centralize cost and model extraction logic in Python and update video project configuration.
+- Update Stage1ProjectSetup for StagePanelProps compatibility, correct annotation fetching API parameter, and switch to 'server-only' directive for database utilities.
+
 ## v0.0.157 (2026-02-22)
 
 ### Feat
 
-- Implement command timeouts in sync regression and add retry logic for prerequisite generation in cloud regression.
-- Implement smooth quadratic bezier curves for waypoint edges and enhance graph layout with wire-length minimization, column reordering, and gravity refinement passes.
-- Introduce waypoint edges for improved dependency graph rendering and ensure backend layout operations are non-mutating.
-- Enhance graph layout with new column-split rules and ensure proper discard of rearranged positions.
-- Implement frontend job management with `useJobs` hook and `JobDashboard` component, alongside prompt file path refactoring.
-- Introduce a comprehensive set of prompts for generating frontend components, hooks, and backend commands, and update related frontend files and configuration.
-- Implement agentic graph layout rearrangement using an LLM to automatically position architecture modules.
-- Implement prompt refresh in ArchitectureView on sync completion and integrate Vitest for frontend testing.
+- **Deadline-aware agentic retry budgeting** — `run_agentic_task` now accepts a `deadline` parameter (also read from `PDD_JOB_DEADLINE` env var). Each retry attempt checks remaining budget against `JOB_TIMEOUT_MARGIN_SECONDS` and `MIN_ATTEMPT_TIMEOUT_SECONDS`, skipping attempts that cannot finish in time. `JobManager` sets a 30-minute job timeout and propagates the deadline to subprocess jobs, with graceful SIGTERM→SIGKILL escalation and zombie process detection.
+- **Agentic graph layout rearrangement** — New `/api/v1/architecture/rearrange` endpoint runs `arrange_graph_layout_LLM.prompt` via `run_agentic_task` to automatically assign swimlane positions to architecture modules. Changes are returned to the frontend without persisting to disk until explicitly saved.
+- **Waypoint edges and improved graph rendering** — Dependency graph now renders smooth quadratic bezier curves through waypoint nodes, with wire-length minimization, column reordering, gravity refinement passes, and new column-split rules. Backend layout operations are non-mutating (original positions preserved for discard).
+- **Frontend job management** — New `useJobs` hook and `JobDashboard` component for monitoring server-side job execution. Prompt refresh triggers automatically on sync completion via `syncCompletedCounter`.
+- **`get_language_outputs()` for config-type language detection** — New function in `construct_paths.py` distinguishes config/data/markup languages (JSON, YAML, CSS, etc.) that produce only `code` from executable languages that produce `code`, `test`, and `example`. Uses the `outputs` column from `language_format.csv` with a built-in fallback set. `moduleNeedsSync` in the frontend uses `expected_outputs` to avoid counting config modules as needing sync.
+- **Comprehensive frontend prompt coverage** — Added 60+ prompts covering all frontend components, hooks, pages, and backend command modules (`checkup`, `firecrawl`, `report`, `which`), plus `config_resolution`, `remote_session`, `generate_model_catalog`, `pin_example_hack`, and `template_expander`.
+- **Vitest integration** — Frontend testing now supports Vitest alongside Jest, with 5 new test suites: waypoint edges, rearrange positions, rearrange discard, sync-completed refresh, and module-needs-sync.
+- **Command timeouts in sync regression** — `run_pdd_command_base` now wraps commands with `PDD_CMD_TIMEOUT` (default 600s). Cloud regression adds retry logic for prerequisite generation failures.
 - Increase LLM call timeout and add project dependencies CSV.
-- Implement the video editor frontend application structure, including Stages 9 and 10, and add deadline propagation to agentic verification.
-- introduce multi-stage video editing workflow with components, examples, and tests for each stage.
-- Implement `VideoPlayer` component with rich annotation capabilities and add `StageSidebar` component, including examples and tests.
-- Add API endpoints for batch annotation resolution and video serving, introduce an SSE log panel component, and update related configuration and job processing.
-- Add annotation API routes for management and analysis, including tests and examples, and update LLM model ELO scores.
 
 ### Fix
 
-- Improve job timeout handling by logging potential zombie processes, add a corresponding test, and enhance error logging and progress reporting for section rendering.
-- Sync Remaining count no longer includes config-type dev units (JSON, YAML, CSS)
-- remove inaccurate RichLog max_lines claim from sync_tui prompt
-- add ADC fallback for VERTEX_CREDENTIALS in Cloud Build CI
-- save RunReport after agentic verify to prevent false crash-verify cycle
-- Correct ELO scores for codex models and expand model catalog
+- **Job timeout handling** — Jobs now enforce a 30-minute wall-clock timeout with SIGTERM→SIGKILL escalation, zombie process logging, and signal-killed process detection (`exit_code < 0`).
+- **Sync Remaining count** — No longer includes config-type dev units (JSON, YAML, CSS) that only produce code output.
+- **RunReport saved after agentic verify** — Prevents false crash-verify cycle where a successful agentic verify was not recorded, causing the state machine to re-enter the fix loop.
+- **Rich markup escaping in error messages** — `fix_verification_errors_loop` now escapes file paths and error messages with `rich_escape()` to prevent Rich markup injection from unusual filenames or error strings.
+- **ADC fallback for Vertex AI** — `_ensure_api_key()` now falls back to Application Default Credentials when `VERTEX_CREDENTIALS` is unset but a GCP project env var is available.
+- Remove inaccurate RichLog `max_lines` claim from `sync_tui` prompt.
+- **Updated ELO scores** — Refreshed from LMArena Code Arena (Feb 22, 2026). Notable changes: `claude-sonnet-4.6` jumps to #3 (1524), new entries for `gpt-5.2-codex`, `gpt-5.1-codex-mini/max`, `gpt-5.3-codex`, `gpt-5-nano`, `gemini-3.1-pro`, and GitHub Copilot naming aliases.
 
 ### Refactor
 
-- update architecture diagram component positions and replace Unicode escape sequences with actual characters."
-- apply custom theme to console output and adjust TypeScript import paths after removing baseUrl.
+- **`_echo_rich` for CliRunner-compatible Rich output** — `templates.py` now renders Rich objects through an in-memory `Console` and outputs via `click.echo`, fixing test capture under `CliRunner`.
+- Update architecture diagram component positions and replace Unicode escape sequences with actual characters.
+- Apply custom theme to console output and adjust TypeScript import paths after removing `baseUrl`.
 
 ## v0.0.156 (2026-02-21)
 

README.md

@@ -1,6 +1,6 @@
 # PDD (Prompt-Driven Development) Command Line Interface
 
-![PDD-CLI Version](https://img.shields.io/badge/pdd--cli-v0.0.157-blue) [![Discord](https://img.shields.io/badge/Discord-join%20chat-7289DA.svg?logo=discord&logoColor=white)](https://discord.gg/Yp4RTh8bG7)
+![PDD-CLI Version](https://img.shields.io/badge/pdd--cli-v0.0.158-blue) [![Discord](https://img.shields.io/badge/Discord-join%20chat-7289DA.svg?logo=discord&logoColor=white)](https://discord.gg/Yp4RTh8bG7)
 
 ## Introduction
 
@@ -352,7 +352,7 @@ For proper model identifiers to use in your custom configuration, refer to the [
 
 ## Version
 
-Current version: 0.0.157
+Current version: 0.0.158
 
 To check your installed version, run:
 ```

examples/hello/repo_root

@@ -1 +1 @@
-Subproject commit dff3772e2073ec0a14de9c24062d1d9d44a16032
+Subproject commit ff62a8bfdec28c16dc4bee0f9ebb6f325b24bc3f

pdd/__init__.py

@@ -2,7 +2,7 @@
 
 import os
 
-__version__ = "0.0.157"
+__version__ = "0.0.158"
 
 # Strength parameter used for LLM extraction across the codebase
 # Used in postprocessing, XML tagging, code generation, and other extraction

pdd/llm_invoke.py

@@ -1210,14 +1210,22 @@ def _ensure_api_key(model_info: Dict[str, Any], newly_acquired_keys: Dict[str, b
             newly_acquired_keys[api_key_field] = False
             return True
 
-        # Vertex AI ADC fallback: GOOGLE_APPLICATION_CREDENTIALS may be unset
-        # if the user ran ``gcloud auth application-default login`` instead.
-        if "GOOGLE_APPLICATION_CREDENTIALS" in env_vars and "GOOGLE_APPLICATION_CREDENTIALS" in missing:
+        # Vertex AI fallback: resolve missing VERTEXAI_PROJECT from
+        # GOOGLE_CLOUD_PROJECT and missing VERTEXAI_LOCATION from CSV location column
+        # (the invocation code already reads CSV location).
+        if "GOOGLE_APPLICATION_CREDENTIALS" in env_vars:
             project = os.getenv("VERTEXAI_PROJECT") or os.getenv("GOOGLE_CLOUD_PROJECT")
-            if project:
-                remaining = [v for v in missing if v != "GOOGLE_APPLICATION_CREDENTIALS"]
+            has_location = (
+                "VERTEXAI_LOCATION" not in missing
+                or bool(model_info.get("location"))
+            )
+            if project and has_location:
+                remaining = [
+                    v for v in missing
+                    if v not in ("GOOGLE_APPLICATION_CREDENTIALS", "VERTEXAI_PROJECT", "VERTEXAI_LOCATION")
+                ]
                 if not remaining:
-                    logger.info(f"Using ADC for Vertex AI (project={project}).")
+                    logger.info(f"Using Vertex AI credentials (project={project}).")
                     newly_acquired_keys[api_key_field] = False
                     return True
 
@@ -2000,6 +2008,11 @@ def calc_strength(candidate):
                 if verbose:
                     logger.info(f"[INFO] No API key for '{model_name_litellm}'; using device flow or default auth.")
 
+            # Pass vertex_location from CSV (e.g., "global" for gemini-3-flash-preview)
+            location = model_info.get('location')
+            if pd.notna(location) and location:
+                litellm_kwargs["vertex_location"] = str(location)
+
             # Add base_url/api_base override if present in CSV
             api_base = model_info.get('base_url')
             if pd.notna(api_base) and api_base:

pdd/preprocess.py

@@ -15,6 +15,10 @@
 install()
 console = Console()
 
+# Maximum iterations for the non-recursive convergence loop.
+# Diamond includes converge quickly; true cycles never converge.
+_MAX_INCLUDE_ITERATIONS = 50
+
 # Debug/Instrumentation controls
 _DEBUG_PREPROCESS = str(os.getenv("PDD_PREPROCESS_DEBUG", "")).lower() in ("1", "true", "yes", "on")
 _DEBUG_OUTPUT_FILE = os.getenv("PDD_PREPROCESS_DEBUG_FILE")  # Optional path to write a debug report
@@ -197,30 +201,15 @@ def replace_include(match):
             console.print(f"[bold red]Error processing include:[/bold red] {str(e)}")
             _dbg(f"Error processing backtick include {file_path}: {e}")
             return f"```[Error processing include: {file_path}]```"
-    _expanded_prior = set()
-    _expanded_current = set()
-
-    def replace_include_tracked(match):
-        file_path = match.group(1).strip()
-        try:
-            full_path = get_file_path(file_path)
-            resolved = os.path.realpath(full_path)
-            if resolved in _expanded_prior:
-                raise ValueError(f"Circular include detected: {file_path} is already in the include chain")
-            _expanded_current.add(resolved)
-        except (FileNotFoundError, ValueError):
-            raise
-        except Exception:
-            pass
-        return replace_include(match)
-
     prev_text = ""
     current_text = text
+    iterations = 0
     while prev_text != current_text:
+        if iterations >= _MAX_INCLUDE_ITERATIONS:
+            raise ValueError("Circular include detected: maximum include depth exceeded")
         prev_text = current_text
-        _expanded_current = set()
-        current_text = re.sub(pattern, replace_include_tracked, current_text, flags=re.DOTALL)
-        _expanded_prior.update(_expanded_current)
+        current_text = re.sub(pattern, replace_include, current_text, flags=re.DOTALL)
+        iterations += 1
     return current_text
 
 def process_xml_tags(text: str, recursive: bool, _seen: Optional[set] = None) -> str:
@@ -309,35 +298,20 @@ def replace_include(match):
             console.print(f"[bold red]Error processing include:[/bold red] {str(e)}")
             _dbg(f"Error processing XML include {file_path}: {e}")
             return f"[Error processing include: {file_path}]"
-    _expanded_prior = set()
-    _expanded_current = set()
-
-    def replace_include_tracked(match):
-        file_path = match.group(1).strip()
-        try:
-            full_path = get_file_path(file_path)
-            resolved = os.path.realpath(full_path)
-            if resolved in _expanded_prior:
-                raise ValueError(f"Circular include detected: {file_path} is already in the include chain")
-            _expanded_current.add(resolved)
-        except (FileNotFoundError, ValueError):
-            raise
-        except Exception:
-            pass
-        return replace_include(match)
-
     prev_text = ""
     current_text = text
+    iterations = 0
     while prev_text != current_text:
+        if iterations >= _MAX_INCLUDE_ITERATIONS:
+            raise ValueError("Circular include detected: maximum include depth exceeded")
         prev_text = current_text
-        _expanded_current = set()
         code_spans = _extract_code_spans(current_text)
         def replace_include_with_spans(match):
             if _intersects_any_span(match.start(), match.end(), code_spans):
                 return match.group(0)
-            return replace_include_tracked(match)
+            return replace_include(match)
         current_text = re.sub(pattern, replace_include_with_spans, current_text, flags=re.DOTALL)
-        _expanded_prior.update(_expanded_current)
+        iterations += 1
     return current_text
 
 def process_pdd_tags(text: str) -> str:

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pdd-cli"
-version = "0.0.157"
+version = "0.0.158"
 description = "PDD (Prompt-Driven Development) Command Line Interface"
 readme = {file = "pypi_description.rst", content-type = "text/x-rst"}
 authors = [
@@ -127,7 +127,7 @@ filterwarnings = [
 
 [tool.commitizen]
 name = "cz_conventional_commits"
-version = "0.0.157"
+version = "0.0.158"
 tag_format = "v$version"
 version_files = [
     "pyproject.toml:version",