tasks-prompts-chain 0.1.0__py3-none-any.whl → 0.1.1__py3-none-any.whl

tasks_prompts_chain/tasks_prompts_chain.py

@@ -31,6 +31,7 @@ Copyright 2025 Samir Ben Sghaier - Smirfolio
 from typing import List, Optional, Dict, Union, AsyncGenerator, TypedDict
 from enum import Enum
 from .client_llm_sdk import ClientLLMSDK
+import json
 
 class OutputFormat(Enum):
     JSON = "JSON"
@@ -46,10 +47,12 @@ class ModelOptions(TypedDict, total=False):
     max_tokens: Optional[int]
 
 class PromptTemplate:
-    def __init__(self, prompt: str, output_format: str = "TEXT", output_placeholder: Optional[str] = None):
+    def __init__(self, prompt: str, output_format: str = "TEXT", output_placeholder: Optional[str] = None, llm_id: Optional[str] = None, stop_placeholder: Optional[str] = None):
         self.prompt = prompt
         self.output_format = OutputFormat(output_format.upper())
         self.output_placeholder = output_placeholder
+        self.llm_id=llm_id
+        self.stop_placeholder = stop_placeholder
 
 class TasksPromptsChain:
     """A utility class for creating and executing prompt chains using OpenAI's API."""
@@ -165,7 +168,8 @@ class TasksPromptsChain:
                                                             "prompt": str,
                                                             "output_format": str,
                                                             "output_placeholder": str,
-                                                            "llm_id": str  # Optional: Specifies which LLM to use
+                                                            "llm_id": str,  # Optional: Specifies which LLM to use
+                                                            "stop_placeholder": str  # Optional: The stop string placeholder
                                                         }
         Returns:
             AsyncGenerator[str, None]: Generator yielding response chunks
@@ -178,24 +182,22 @@ class TasksPromptsChain:
                 # Convert dict to PromptTemplate if necessary and extract llm_id
                 llm_id = None
                 if isinstance(prompt_data, dict):
-                    # Extract llm_id from the prompt data if present
-                    llm_id = prompt_data.get("llm_id", self.default_client_id)
                     prompt_template = PromptTemplate(
                         prompt=prompt_data["prompt"],
                         output_format=prompt_data.get("output_format", "TEXT"),
-                        output_placeholder=prompt_data.get("output_placeholder")
+                        output_placeholder=prompt_data.get("output_placeholder"),
+                        llm_id= prompt_data.get("llm_id", self.default_client_id),
+                        stop_placeholder=prompt_data.get("stop_placeholder", None)
                     )
                 else:
                     prompt_template = prompt_data
-                    # Use default client if llm_id not specified
-                    llm_id = self.default_client_id
                 
                 # Validate the requested LLM exists
-                if llm_id not in self.clients:
-                    raise ValueError(f"LLM with id '{llm_id}' not found. Available LLMs: {list(self.clients.keys())}")
+                if prompt_template.llm_id not in self.clients:
+                    raise ValueError(f"LLM with id '{prompt_template.llm_id}' not found. Available LLMs: {list(self.clients.keys())}")
                 
                 # Get the client configuration
-                client_config = self.clients[llm_id]
+                client_config = self.clients[prompt_template.llm_id]
                 client = client_config["client"]
                 model = client_config["model"]
                 temperature = client_config["temperature"]
@@ -241,7 +243,13 @@ class TasksPromptsChain:
                             placeholder_values[prompt_template.output_placeholder] = response_content
                             self._results[prompt_template.output_placeholder] = response_content
                         if streamout:
-                            yield delta            
+                            yield delta
+                
+                # Stop excution if the stop_placeholder is detected in the response content
+                if prompt_template.stop_placeholder and (prompt_template.stop_placeholder in response_content):
+                    raise Exception({"type": "error", "content": "Invalid project description. Please provide a valid project description."})
+                    #yield json.dumps({"type": "error", "content": "Invalid project description. Please provide a valid project description."})
+                    #return          
 
         except Exception as e:
             raise Exception(f"Error in prompt chain execution at prompt {i}: {str(e)}")
@@ -285,4 +293,3 @@ class TasksPromptsChain:
         if len(self._results) > 0:
             raise Exception("template_output must be called before execute_chain")
         self.set_output_template(template)
-

{tasks_prompts_chain-0.1.0.dist-info → tasks_prompts_chain-0.1.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tasks_prompts_chain
-Version: 0.1.0
+Version: 0.1.1
 Summary: A Python library for creating and executing chains of prompts using multiple LLM providers with streaming support and template formatting.
 Project-URL: Homepage, https://github.com/smirfolio/tasks_prompts_chain
 Project-URL: Issues, https://github.com/smirfolio/tasks_prompts_chain/issues
@@ -23,6 +23,7 @@ A Mini Python library for creating and executing chains of prompts using multipl
 - Template-based output formatting
 - System prompt support
 - Placeholder replacement between prompts
+- Stop placeholder string, that can be defined, so if the LLM responds with this placeholder, the chain prompt will interrupt, as an LLM error handler
 - Multiple output formats (JSON, Markdown, CSV, Text)
 - Async/await support
 - Support for multiple LLM providers (OpenAI, Anthropic, Cerebras, etc.)
@@ -120,10 +121,11 @@ async def main():
     # Define your prompts - specify which LLM to use for each prompt
     prompts = [
         {
-            "prompt": "Create a design concept for a luxury chocolate bar",
+            "prompt": "Create a design concept for a luxury chocolate bar, if not inspired respond with this string %%cant_be_inspired%% so the chain prompt query will be stopped",
             "output_format": "TEXT",
             "output_placeholder": "design_concept",
-            "llm_id": "gpt"  # Use the GPT model for this prompt
+            "llm_id": "gpt",  # Use the GPT model for this prompt
+            "stop_placholder": "%%cant_be_inspired%%" # the stop placeholder string that will interrupt the prompt chain query
         },
         {
             "prompt": "Based on this concept: {{design_concept}}, suggest a color palette",
@@ -290,7 +292,8 @@ Each prompt in the chain can be defined as a dictionary:
     "prompt": str,              # The actual prompt text
     "output_format": str,       # "JSON", "MARKDOWN", "CSV", or "TEXT"
     "output_placeholder": str,  # Identifier for accessing this result
-    "llm_id": str               # Optional: ID of the LLM to use for this prompt
+    "llm_id": str,              # Optional: ID of the LLM to use for this prompt
+    "stop_placholder": str      # Optional: The stop string placeholder that may interrupt the chaining prompt query, defined in a prompt, and may be returned by the LLM
 }
 ```
 
@@ -311,6 +314,7 @@ The library includes comprehensive error handling:
 - API error handling
 - Placeholder validation
 - LLM validation (checks if specified LLM ID exists)
+- stop_placholder to validate the LLM output and stop the chain prompt excution
 
 Errors are raised with descriptive messages indicating the specific issue and prompt number where the error occurred.
 
@@ -318,6 +322,7 @@ Errors are raised with descriptive messages indicating the specific issue and pr
 
 1. Always set templates before executing the chain
 2. Use meaningful placeholder names
+3. Implement a stop_placholder in each prompt, to catch the LLM error defined bad responses to stop the subsequent requests
 3. Handle streaming responses appropriately
 4. Choose appropriate models for different types of tasks
 5. Use system prompts for consistent context

tasks_prompts_chain-0.1.1.dist-info/RECORD

@@ -0,0 +1,7 @@
+tasks_prompts_chain/__init__.py,sha256=HVhC_vMTYCyZW6vnoErHh-TkAnNRqJ2JJqClJQSfU8Y,148
+tasks_prompts_chain/client_llm_sdk.py,sha256=ifwsecvykP5RalXMqolAClPHdpmv-5hF4eT61dxo1f8,3708
+tasks_prompts_chain/tasks_prompts_chain.py,sha256=T8WaqAdgDBqHjNa48dPgny_mDnEN1OvMTFTX6YndGgo,13409
+tasks_prompts_chain-0.1.1.dist-info/METADATA,sha256=1zVlU2deVaP-Fv_Zj6ph6LcDJrJOD4qSffehnp1umRM,13331
+tasks_prompts_chain-0.1.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+tasks_prompts_chain-0.1.1.dist-info/licenses/LICENSE,sha256=WYmcYJG1QFgu1hfo7qrEkZ3Jhcz8NUWe6XUraZvlIFs,10172
+tasks_prompts_chain-0.1.1.dist-info/RECORD,,

tasks_prompts_chain-0.1.0.dist-info/RECORD

@@ -1,7 +0,0 @@
-tasks_prompts_chain/__init__.py,sha256=HVhC_vMTYCyZW6vnoErHh-TkAnNRqJ2JJqClJQSfU8Y,148
-tasks_prompts_chain/client_llm_sdk.py,sha256=ifwsecvykP5RalXMqolAClPHdpmv-5hF4eT61dxo1f8,3708
-tasks_prompts_chain/tasks_prompts_chain.py,sha256=ZqZIzAWEZMgKvbkXUT8_VZisSydb-jLuWxuTrMT1-Fw,12643
-tasks_prompts_chain-0.1.0.dist-info/METADATA,sha256=_-5r7zd9EnnPut3CF45s1VFGVte0iZHYb9Epx219UfQ,12573
-tasks_prompts_chain-0.1.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-tasks_prompts_chain-0.1.0.dist-info/licenses/LICENSE,sha256=WYmcYJG1QFgu1hfo7qrEkZ3Jhcz8NUWe6XUraZvlIFs,10172
-tasks_prompts_chain-0.1.0.dist-info/RECORD,,