Bump version

Author: gltanaka

CHANGELOG.md

@@ -1,17 +1,25 @@
+## v0.0.156 (2026-02-21)
+
+### Feat
+
+- Add new API routes for annotations, job details, and job retry, with corresponding examples and integration tests.
+- Implement core video editor pipeline APIs, scripts, tests, and examples for project, audio, TTS, and rendering functionalities.
+- Implement pagination for `RemoteSessionManager.list_sessions` a… (#492)
+
+### Fix
+
+- Correct log line processing in streaming routes by excluding trailing empty strings and include 'logs' in job schema tests.
+
 ## v0.0.155 (2026-02-20)
 
 ### Feat
 
-- Add initial video editor project scaffolding with prompts, Dockerfile, and configuration, and update CI test durations.
-- initialize video editor project with core files, prompts, documentation, and CI configuration.
-- Add support for generating a video editor application with its architecture, prompts, and initial structure.
-- replace placeholder script with a detailed 'Prompt-Driven Development' video script for the narrative demo.
-- Introduce `pdd checkup` command, update models, refactor `.pddrc` strategy detection, migrate package tests to Cloud Build, and update grounding experiment documentation.
-- Dry-run validation with per-module cwd for agentic sync (#489)
+- **Dry-run validation with per-module cwd in agentic sync** — Before dispatching parallel `pdd sync` workers, `run_agentic_sync` now runs a dry-run validation phase for every module. `_resolve_module_cwd()` discovers the correct working directory by scanning `.pddrc` files at the project root and up to two levels of subdirectories; the deepest matching context wins. `_run_single_dry_run()` executes `pdd sync <basename> --dry-run --agentic --no-steer` from the resolved cwd. If the dry-run fails, `_llm_fix_dry_run_failure()` builds a project-tree snapshot and `.pddrc` location list, invokes the LLM via a new `agentic_sync_fix_dry_run_LLM.prompt`, extracts the `SYNC_CWD:` marker from the response, validates the suggested path is inside the project root, and re-runs the dry-run to confirm. The resulting `module_cwds` dict is passed to `AsyncSyncRunner`, which now uses a per-module cwd instead of always running from the project root.
+- **Subproject directory support in agentic architecture** — `pdd generate --agentic` and the underlying orchestrator now accept a `--output-dir` / `target_dir` option to scope a new project to a subdirectory. `_extract_target_dir()` auto-parses the target directory from the GitHub issue body using tight patterns (backtick/double-quote wrapped, or unquoted paths containing `_` or `/` to avoid false matches on natural English). The orchestrator derives `base_dir = cwd / target_dir` and creates it if needed; all `architecture.json` reads/writes and scaffolding file creation are redirected to `base_dir`. Step 6 gains an MD5 hash check before and after execution — if the file content is unchanged the step is marked failed immediately, preventing silent no-ops where the agent reformats the existing file instead of creating a new architecture. The `agentic_arch_step6_generate_LLM.prompt` now includes explicit `target_dir` instructions.
 
 ### Fix
 
-- improve e2e test resilience by skipping on LLM generation failures or empty responses.
+- **E2E test resilience** — `test_e2e_issue_342_syspath_isolation` now skips (instead of failing) when the LLM returns an empty response or a generation failure, making CI stable in environments without an API key.
 
 ## v0.0.154 (2026-02-19)
 

README.md

@@ -1,6 +1,6 @@
 # PDD (Prompt-Driven Development) Command Line Interface
 
-![PDD-CLI Version](https://img.shields.io/badge/pdd--cli-v0.0.155-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-v0.0.156-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: 0.0.155
+Current version: 0.0.156
 
 To check your installed version, run:
 ```

docs/prompting_guide.md

@@ -195,10 +195,10 @@ Tip: Prefer small, named sections using XML‑style tags to make context scannab
 
 The PDD preprocessor supports additional XML‑style tags to keep prompts clean, reproducible, and self‑contained. Processing order (per spec) is: `pdd` → `include`/`include-many` → `shell` → `web`. When `recursive=True`, `<shell>` and `<web>` are deferred until a non‑recursive pass.
 
-- `<pdd>…</pdd>`
+- ``
   - Purpose: human‑only comment. Removed entirely during preprocessing.
   - Use: inline rationale or notes that should not reach the model.
-  - Example: `Before step X <pdd>explain why we do this here</pdd>`
+  - Example: `Before step X `
 
 - `<shell>…</shell>`
   - Purpose: run a shell command and inline stdout at that position.
@@ -259,7 +259,7 @@ Place architecture metadata tags at the **top of your prompt file** (after any `
   "type": "module",
   "module": {{
     "functions": [
-      {{"name": "function_name", "signature": "(...)", "returns": "Type"}}
+      {"name": "function_name", "signature": "(...)", "returns": "Type"}
     ]
   }}
 }}
@@ -288,7 +288,7 @@ Place architecture metadata tags at the **top of your prompt file** (after any `
     "type": "module",
     "module": {{
       "functions": [
-        {{"name": "llm_invoke", "signature": "(prompt, strength, ...)", "returns": "Dict"}}
+        {"name": "llm_invoke", "signature": "(prompt, strength, ...)", "returns": "Dict"}
       ]
     }}
   }}
@@ -906,4 +906,4 @@ Key differences:
 
 ## Final Notes
 
-Think of prompts as your programming language. Keep them concise, explicit, and modular. Regenerate instead of patching, verify behavior with accumulating tests, and continuously back‑propagate implementation learnings into your prompts. That discipline is what converts maintenance from an endless patchwork into a compounding system of leverage.
+Think of prompts as your programming language. Keep them concise, explicit, and modular. Regenerate instead of patching, verify behavior with accumulating tests, and continuously back‑propagate implementation learnings into your prompts. That discipline is what converts maintenance from an endless patchwork into a compounding system of leverage.

examples/hello/repo_root

@@ -1 +1 @@
-Subproject commit 1d2a807b4cb44896484f5f1ff78255b8108b7398
+Subproject commit 315330d11167fa3a9370d39286f871913b25e51a

pdd/__init__.py

@@ -2,7 +2,7 @@
 
 import os
 
-__version__ = "0.0.155"
+__version__ = "0.0.156"
 
 # Strength parameter used for LLM extraction across the codebase
 # Used in postprocessing, XML tagging, code generation, and other extraction

pdd/commands/modify.py

@@ -213,6 +213,19 @@ def update(
     Single-file mode (1 arg): Update prompt for specific code file.
     """
     ctx.ensure_object(dict)
+
+    # Validate argument counts before try/except so UsageError propagates naturally
+    if len(files) == 2 and not git:
+        raise click.UsageError(
+            "Two arguments require --git flag: pdd update --git <prompt> <modified_code>"
+        )
+    if len(files) == 3 and git:
+        raise click.UsageError(
+            "Cannot use --git with 3 arguments (--git and original_code are mutually exclusive)"
+        )
+    if len(files) > 3:
+        raise click.UsageError("Too many arguments. Max 3: <prompt> <modified_code> <original_code>")
+
     try:
         # Handle argument counts per modify_python.prompt spec (aligned with README)
         if len(files) == 0:
@@ -228,27 +241,17 @@ def update(
             modified_code_file = files[0]
             input_code_file = None
         elif len(files) == 2:
-            # Git-based update: prompt + modified_code (requires --git)
-            if not git:
-                raise click.UsageError(
-                    "Two arguments require --git flag: pdd update --git <prompt> <modified_code>"
-                )
+            # Git-based update: prompt + modified_code (--git guaranteed by pre-validation)
             is_repo_mode = False
             input_prompt_file = files[0]
             modified_code_file = files[1]
             input_code_file = None
         elif len(files) == 3:
-            # Manual update: prompt + modified_code + original_code
-            if git:
-                raise click.UsageError(
-                    "Cannot use --git with 3 arguments (--git and original_code are mutually exclusive)"
-                )
+            # Manual update: prompt + modified_code + original_code (no --git guaranteed)
             is_repo_mode = False
             input_prompt_file = files[0]
             modified_code_file = files[1]
             input_code_file = files[2]
-        else:
-            raise click.UsageError("Too many arguments. Max 3: <prompt> <modified_code> <original_code>")
 
         # Validate mode-specific options
         if is_repo_mode:
@@ -288,7 +291,7 @@ def update(
 
         return result, cost, model
 
-    except click.Abort:
+    except (click.Abort, click.UsageError):
         raise
     except Exception as e:
         handle_error(e, "update", ctx.obj.get("quiet", False))

pdd/core/cli.py

@@ -108,6 +108,8 @@ def invoke(self, ctx):
             # User cancelled (e.g., pressed 'no' on confirmation) - set flag
             # to exit silently without triggering error reporting
             user_abort = True
+        except click.UsageError:
+            raise  # Let Click handle it natively via BaseCommand.main() → echo()
         except KeyboardInterrupt as e:
             # Handle keyboard interrupt (Ctrl+C) gracefully
             exception_to_handle = e