pdd-cli 0.0.23__py3-none-any.whl → 0.0.25__py3-none-any.whl

pdd/llm_invoke.py

@@ -14,7 +14,7 @@ Usage:
     # result is a dict with keys: 'result', 'cost', 'model_name'
     
 Environment:
-    - PDD_MODEL_DEFAULT: if set, used as the base model name. Otherwise defaults to "gpt-4o-mini".
+    - PDD_MODEL_DEFAULT: if set, used as the base model name. Otherwise defaults to "gpt-4.1-nano".
     - PDD_PATH: if set, models are loaded from $PDD_PATH/data/llm_model.csv; otherwise from ./data/llm_model.csv.
     - Models that require an API key will check the corresponding environment variable (name provided in the CSV).
 """
@@ -25,6 +25,7 @@ import json
 
 from pydantic import BaseModel, Field
 from rich import print as rprint
+from rich.errors import MarkupError
 
 # Langchain core and community imports
 from langchain_core.prompts import PromptTemplate, ChatPromptTemplate
@@ -218,10 +219,10 @@ def create_llm_instance(selected_model, temperature, handler):
                              openai_api_key=api_key, callbacks=[handler],
                              openai_api_base=base_url)
         else:
-            if model_name.startswith('o') and 'mini' not in model_name:
+            if model_name.startswith('o'):
                 llm = ChatOpenAI(model=model_name, temperature=temperature,
                                  openai_api_key=api_key, callbacks=[handler],
-                                 reasoning_effort='high')
+                                 reasoning={"effort": "high","summary": "auto"})
             else:
                 llm = ChatOpenAI(model=model_name, temperature=temperature,
                                  openai_api_key=api_key, callbacks=[handler])
@@ -301,7 +302,7 @@ def llm_invoke(prompt, input_json, strength, temperature, verbose=False, output_
         raise ValueError("Input JSON must be a dictionary.")
 
     set_llm_cache(SQLiteCache(database_path=".langchain.db"))
-    base_model_name = os.environ.get('PDD_MODEL_DEFAULT', 'gpt-4o-mini')
+    base_model_name = os.environ.get('PDD_MODEL_DEFAULT', 'gpt-4.1-nano')
     models = load_models()
     
     try:
@@ -326,7 +327,6 @@ def llm_invoke(prompt, input_json, strength, temperature, verbose=False, output_
             llm = create_llm_instance(model, temperature, handler)
             if output_pydantic:
                 if model.structured_output:
-                    llm.cache = False # TODO: remove this fix once langchain cache is fixed https://github.com/langchain-ai/langchain/issues/29003
                     llm = llm.with_structured_output(output_pydantic)
                     chain = prompt_template | llm
                 else:
@@ -348,12 +348,22 @@ def llm_invoke(prompt, input_json, strength, temperature, verbose=False, output_
                 rprint(f"Strength used: {strength}")
                 rprint(f"Temperature used: {temperature}")
                 try:
-                    rprint(f"Input JSON: {str(input_json)}")  # Use str() instead of json.dumps()
+                    # Try printing with rich formatting first
+                    rprint(f"Input JSON: {str(input_json)}")
+                except MarkupError:
+                    # Fallback to standard print if rich markup fails
+                    print(f"Input JSON: {str(input_json)}")
                 except Exception:
-                    rprint(f"Input JSON: {input_json}")
+                    print(f"Input JSON: {input_json}")
                 if output_pydantic:
                     rprint(f"Output Pydantic format: {output_pydantic}")
-                rprint(f"Result: {result_output}")
+                try:
+                    # Try printing with rich formatting first
+                    rprint(f"Result: {result_output}")
+                except MarkupError as me:
+                    # Fallback to standard print if rich markup fails
+                    print(f"[bold yellow]Warning:[/bold yellow] Failed to render result with rich markup: {me}")
+                    print(f"Raw Result: {str(result_output)}") # Use standard print
 
             return {'result': result_output, 'cost': cost, 'model_name': model.model}
 

pdd/pdd_completion.zsh

@@ -1,5 +1,35 @@
 #compdef pdd
 
+##
+# ZSH Completion for PDD CLI (Prompt-Driven Development)
+#
+# Save this file as "pdd_completion.zsh" and source it from your ~/.zshrc:
+#   source /path/to/pdd_completion.zsh
+#
+# The script will handle completion initialization automatically.
+#
+# After installation, typing:
+#   pdd <Tab>
+# will offer completions for subcommands and options as described in the PDD CLI README.
+##
+
+# First, make sure we're using zsh
+if [ -z "$ZSH_VERSION" ]; then
+   echo >&2 "pdd completion requires zsh."
+   return 1
+fi
+
+# Add this directory to fpath so ZSH can find our completion function
+script_dir=${0:A:h}
+fpath=($script_dir $fpath)
+
+# Check if we need to initialize completion system
+# Use command -v to check if compdef function exists
+if ! command -v compdef >/dev/null 2>&1; then
+    autoload -U compinit
+    compinit
+fi
+
 ##
 # ZSH Completion for PDD CLI (Prompt-Driven Development)
 #
@@ -413,6 +443,13 @@ _pdd() {
   esac
 }
 
-compdef _pdd pdd
+# Register the _pdd function as a completion for pdd command
+# Use command -v to safely check if compdef is available again
+# (in case something went wrong with the initialization)
+if command -v compdef >/dev/null 2>&1; then
+  compdef _pdd pdd
+else
+  echo >&2 "Warning: Could not register pdd completion. Make sure ZSH completion system is working."
+fi
 
 # End of pdd_completion.zsh

pdd/preprocess.py

@@ -133,9 +133,9 @@ def process_web_tags(text: str) -> str:
                 console.print("[bold yellow]Warning:[/bold yellow] FIRECRAWL_API_KEY not found in environment")
                 return f"[Error: FIRECRAWL_API_KEY not set. Cannot scrape {url}]"
             app = FirecrawlApp(api_key=api_key)
-            response = app.scrape_url(url=url, params={'formats': ['markdown']})
-            if 'markdown' in response:
-                return response['markdown']
+            response = app.scrape_url(url, formats=['markdown'])
+            if hasattr(response, 'markdown'):
+                return response.markdown
             else:
                 console.print(f"[bold yellow]Warning:[/bold yellow] No markdown content returned for {url}")
                 return f"[No content available for {url}]"