Rename core/llm_clients.py to core/llm_services.py

core/llm_clients.py

@@ -1,126 +0,0 @@
-# algoforge_prime/core/llm_clients.py
-import os
-import google.generativeai as genai
-from huggingface_hub import InferenceClient
-import time
-
-# --- Configuration ---
-GOOGLE_API_KEY = None
-HF_TOKEN = None
-GEMINI_API_CONFIGURED = False
-HF_API_CONFIGURED = False
-hf_inference_client = None
-
-def initialize_all_clients():
-    global GOOGLE_API_KEY, HF_TOKEN, GEMINI_API_CONFIGURED, HF_API_CONFIGURED, hf_inference_client
-    print("INFO: llm_clients.py - Attempting to initialize all API clients...")
-    GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
-    if GOOGLE_API_KEY and GOOGLE_API_KEY.strip():
-        print("INFO: llm_clients.py - GOOGLE_API_KEY found.")
-        try:
-            genai.configure(api_key=GOOGLE_API_KEY)
-            GEMINI_API_CONFIGURED = True
-            print("SUCCESS: llm_clients.py - Google Gemini API configured.")
-        except Exception as e:
-            GEMINI_API_CONFIGURED = False
-            print(f"ERROR: llm_clients.py - Failed to configure Google Gemini API: {type(e).__name__}: {e}")
-    else:
-        GEMINI_API_CONFIGURED = False
-        print("WARNING: llm_clients.py - GOOGLE_API_KEY not found or empty.")
-
-    HF_TOKEN = os.getenv("HF_TOKEN")
-    if HF_TOKEN and HF_TOKEN.strip():
-        print("INFO: llm_clients.py - HF_TOKEN found.")
-        try:
-            hf_inference_client = InferenceClient(token=HF_TOKEN)
-            HF_API_CONFIGURED = True
-            print("SUCCESS: llm_clients.py - Hugging Face InferenceClient initialized.")
-        except Exception as e:
-            HF_API_CONFIGURED = False
-            print(f"ERROR: llm_clients.py - Failed to initialize HF InferenceClient: {type(e).__name__}: {e}")
-            hf_inference_client = None
-    else:
-        HF_API_CONFIGURED = False
-        print("WARNING: llm_clients.py - HF_TOKEN not found or empty.")
-    print(f"INFO: llm_clients.py - Init complete. Gemini Configured: {GEMINI_API_CONFIGURED}, HF Configured: {HF_API_CONFIGURED}")
-
-# --- Status Getter Functions ---
-def is_gemini_api_configured():
-    global GEMINI_API_CONFIGURED
-    return GEMINI_API_CONFIGURED
-
-def is_hf_api_configured():
-    global HF_API_CONFIGURED
-    return HF_API_CONFIGURED
-
-# ... (LLMResponse class and call_huggingface_api function remain the same as the last full version) ...
-class LLMResponse: # Make sure this is defined
-    def __init__(self, text=None, error=None, success=True, raw_response=None, model_id_used="unknown"):
-        self.text, self.error, self.success, self.raw_response, self.model_id_used = text, error, success, raw_response, model_id_used
-    def __str__(self): return str(self.text) if self.success and self.text is not None else f"ERROR (Model: {self.model_id_used}): {self.error}"
-
-def call_huggingface_api(prompt_text, model_id, temperature=0.7, max_new_tokens=512, system_prompt_text=None):
-    print(f"DEBUG: llm_clients.py - call_huggingface_api for model: {model_id}")
-    if not is_hf_api_configured() or not hf_inference_client: # Use getter
-        error_msg = "HF API not configured."
-        print(f"ERROR: llm_clients.py - {error_msg}")
-        return LLMResponse(error=error_msg, success=False, model_id_used=model_id)
-    full_prompt = f"<s>[INST] <<SYS>>\n{system_prompt_text}\n<</SYS>>\n\n{prompt_text} [/INST]" if system_prompt_text else prompt_text
-    try:
-        print(f"  HF API Call - Prompt (first 100): {full_prompt[:100]}...")
-        use_sample = temperature > 0.001
-        raw_response = hf_inference_client.text_generation(full_prompt, model=model_id, max_new_tokens=max_new_tokens, temperature=temperature if use_sample else None, do_sample=use_sample, stream=False)
-        print(f"  HF API Call - Success for {model_id}. Response (first 100): {str(raw_response)[:100]}...")
-        return LLMResponse(text=raw_response, raw_response=raw_response, model_id_used=model_id)
-    except Exception as e:
-        error_msg = f"HF API Error ({model_id}): {type(e).__name__} - {str(e)}"
-        print(f"ERROR: llm_clients.py - {error_msg}")
-        return LLMResponse(error=error_msg, success=False, raw_response=e, model_id_used=model_id)
-
-
-def call_gemini_api(prompt_text, model_id, temperature=0.7, max_new_tokens=1024, system_prompt_text=None): # Increased default max_tokens
-    print(f"DEBUG: llm_clients.py - call_gemini_api for model: {model_id}")
-    if not is_gemini_api_configured(): # Use getter
-        error_msg = "Google Gemini API not configured."
-        print(f"ERROR: llm_clients.py - {error_msg}")
-        return LLMResponse(error=error_msg, success=False, model_id_used=model_id)
-    try:
-        print(f"  Gemini API Call - Getting instance for: {model_id}")
-        model_instance = genai.GenerativeModel(model_name=model_id, system_instruction=system_prompt_text)
-        generation_config = genai.types.GenerationConfig(temperature=temperature, max_output_tokens=max_new_tokens)
-        print(f"  Gemini API Call - User Prompt (first 100): {prompt_text[:100]}...")
-        if system_prompt_text: print(f"  Gemini API Call - System Prompt (first 100): {system_prompt_text[:100]}...")
-        raw_response = model_instance.generate_content(prompt_text, generation_config=generation_config, stream=False)
-        print(f"  Gemini API Call - Raw response for {model_id}. Feedback: {raw_response.prompt_feedback}, Candidates: {'Yes' if raw_response.candidates else 'No'}")
-
-        if raw_response.prompt_feedback and raw_response.prompt_feedback.block_reason:
-            reason = raw_response.prompt_feedback.block_reason_message or raw_response.prompt_feedback.block_reason
-            error_msg = f"Gemini API: Prompt blocked. Reason: {reason}."
-            print(f"WARNING: llm_clients.py - {error_msg}")
-            return LLMResponse(error=error_msg, success=False, raw_response=raw_response, model_id_used=model_id)
-        if not raw_response.candidates:
-            error_msg = "Gemini API: No candidates in response (often due to blocking)."
-            if raw_response.prompt_feedback: error_msg += f" Feedback: {raw_response.prompt_feedback}"
-            print(f"WARNING: llm_clients.py - {error_msg}")
-            return LLMResponse(error=error_msg, success=False, raw_response=raw_response, model_id_used=model_id)
-
-        candidate = raw_response.candidates[0]
-        if not candidate.content or not candidate.content.parts:
-            finish_reason = str(candidate.finish_reason if candidate.finish_reason else "UNKNOWN").upper()
-            error_msg = f"Gemini API: No content parts. Finish Reason: {finish_reason}."
-            if finish_reason == "SAFETY": error_msg += " Likely safety filters."
-            print(f"WARNING: llm_clients.py - {error_msg}")
-            partial_text = candidate.content.parts[0].text if candidate.content and candidate.content.parts and hasattr(candidate.content.parts[0], 'text') else ""
-            return LLMResponse(text=partial_text + f"\n[Note: Generation ended: {finish_reason}]" if partial_text else None, error=error_msg if not partial_text else None, success=bool(partial_text), raw_response=raw_response, model_id_used=model_id)
-        
-        response_text = candidate.content.parts[0].text
-        print(f"  Gemini API Call - Success for {model_id}. Response (first 100): {response_text[:100]}...")
-        return LLMResponse(text=response_text, raw_response=raw_response, model_id_used=model_id)
-    except Exception as e:
-        error_msg = f"Gemini API Exception ({model_id}): {type(e).__name__} - {str(e)}"
-        # ... (specific error parsing as before) ...
-        if "API key not valid" in str(e) or "PERMISSION_DENIED" in str(e): error_msg = f"Gemini API Error ({model_id}): API key invalid/permission denied. Check GOOGLE_API_KEY & Google Cloud. Original: {str(e)}"
-        elif "Could not find model" in str(e) : error_msg = f"Gemini API Error ({model_id}): Model ID '{model_id}' not found/inaccessible. Original: {str(e)}"
-        elif "Quota exceeded" in str(e): error_msg = f"Gemini API Error ({model_id}): API quota exceeded. Original: {str(e)}"
-        print(f"ERROR: llm_clients.py - {error_msg}")
-        return LLMResponse(error=error_msg, success=False, raw_response=e, model_id_used=model_id)

core/llm_services.py

@@ -0,0 +1,161 @@
+# storyverse_weaver/core/llm_services.py
+import os
+import google.generativeai as genai
+from huggingface_hub import InferenceClient
+# from dotenv import load_dotenv # Optional: for local .env file
+# load_dotenv() # Load environment variables from .env file if present
+
+GOOGLE_API_KEY = os.getenv("STORYVERSE_GOOGLE_API_KEY") # Use specific env var names
+HF_TOKEN = os.getenv("STORYVERSE_HF_TOKEN")
+
+GEMINI_TEXT_CONFIGURED = False
+HF_TEXT_CONFIGURED = False
+hf_inference_text_client = None
+
+class LLMTextResponse:
+    def __init__(self, text=None, error=None, success=True, model_id_used="unknown_text_llm"):
+        self.text, self.error, self.success, self.model_id_used = text, error, success, model_id_used
+
+def initialize_text_llms():
+    global GOOGLE_API_KEY, HF_TOKEN, GEMINI_TEXT_CONFIGURED, HF_TEXT_CONFIGURED, hf_inference_text_client
+    print("INFO: llm_services.py - Initializing Text LLM clients...")
+    if GOOGLE_API_KEY and GOOGLE_API_KEY.strip():
+        try:
+            genai.configure(api_key=GOOGLE_API_KEY)
+            GEMINI_TEXT_CONFIGURED = True
+            print("SUCCESS: llm_services.py - Google Gemini API (for text) configured.")
+        except Exception as e:
+            print(f"ERROR: llm_services.py - Failed to configure Google Gemini API: {e}")
+            GEMINI_TEXT_CONFIGURED = False
+    else:
+        print("WARNING: llm_services.py - STORYVERSE_GOOGLE_API_KEY not found or empty.")
+        GEMINI_TEXT_CONFIGURED = False
+
+    if HF_TOKEN and HF_TOKEN.strip():
+        try:
+            hf_inference_text_client = InferenceClient(token=HF_TOKEN)
+            HF_TEXT_CONFIGURED = True
+            print("SUCCESS: llm_services.py - Hugging Face InferenceClient (for text) initialized.")
+        except Exception as e:
+            print(f"ERROR: llm_services.py - Failed to initialize HF InferenceClient: {e}")
+            HF_TEXT_CONFIGURED = False
+    else:
+        print("WARNING: llm_services.py - STORYVERSE_HF_TOKEN not found or empty.")
+        HF_TEXT_CONFIGURED = False
+    print(f"INFO: llm_services.py - Text LLM Init complete. Gemini Text: {GEMINI_TEXT_CONFIGURED}, HF Text: {HF_TEXT_CONFIGURED}")
+
+def is_gemini_text_ready(): return GEMINI_TEXT_CONFIGURED
+def is_hf_text_ready(): return HF_TEXT_CONFIGURED
+
+def generate_text_gemini(prompt: str, model_id: str = "gemini-1.5-flash-latest", system_prompt: str = None, temperature: float = 0.7, max_tokens: int = 512) -> LLMTextResponse:
+    if not is_gemini_text_ready():
+        return LLMTextResponse(error="Gemini text API not configured.", success=False, model_id_used=model_id)
+    try:
+        model = genai.GenerativeModel(model_name=model_id, system_instruction=system_prompt)
+        config = genai.types.GenerationConfig(temperature=temperature, max_output_tokens=max_tokens)
+        response = model.generate_content(prompt, generation_config=config)
+        # Add robust response checking as in AlgoForge's llm_clients.py
+        if response.prompt_feedback and response.prompt_feedback.block_reason:
+            return LLMTextResponse(error=f"Gemini: Prompt blocked ({response.prompt_feedback.block_reason})", success=False, model_id_used=model_id)
+        if not response.candidates or not response.candidates[0].content.parts:
+            return LLMTextResponse(error=f"Gemini: No content generated (Finish reason: {response.candidates[0].finish_reason if response.candidates else 'Unknown'})", success=False, model_id_used=model_id)
+        return LLMTextResponse(text=response.text, model_id_used=model_id)
+    except Exception as e:
+        return LLMTextResponse(error=f"Gemini API Error ({model_id}): {type(e).__name__} - {str(e)}", success=False, model_id_used=model_id)
+
+def generate_text_hf(prompt: str, model_id: str = "mistralai/Mistral-7B-Instruct-v0.2", system_prompt: str = None, temperature: float = 0.7, max_tokens: int = 512) -> LLMTextResponse:
+    if not is_hf_text_ready() or not hf_inference_text_client:
+        return LLMTextResponse(error="HF text API not configured.", success=False, model_id_used=model_id)
+    
+    full_prompt = f"<s>[INST] <<SYS>>\n{system_prompt}\n<</SYS>>\n\n{prompt} [/INST]" if system_prompt else prompt
+    try:
+        use_sample = temperature > 0.001
+        response_text = hf_inference_text_client.text_generation(
+            full_prompt, model=model_id, max_new_tokens=max_tokens,
+            temperature=temperature if use_sample else None, do_sample=use_sample
+        )
+        return LLMTextResponse(text=response_text, model_id_used=model_id)
+    except Exception as e:
+        return LLMTextResponse(error=f"HF API Error ({model_id}): {type(e).__name__} - {str(e)}", success=False, model_id_used=model_id)
+
+print("DEBUG: core.llm_services (for StoryVerseWeaver) - Module defined.")# storyverse_weaver/core/llm_services.py
+import os
+import google.generativeai as genai
+from huggingface_hub import InferenceClient
+# from dotenv import load_dotenv # Optional: for local .env file
+# load_dotenv() # Load environment variables from .env file if present
+
+GOOGLE_API_KEY = os.getenv("STORYVERSE_GOOGLE_API_KEY") # Use specific env var names
+HF_TOKEN = os.getenv("STORYVERSE_HF_TOKEN")
+
+GEMINI_TEXT_CONFIGURED = False
+HF_TEXT_CONFIGURED = False
+hf_inference_text_client = None
+
+class LLMTextResponse:
+    def __init__(self, text=None, error=None, success=True, model_id_used="unknown_text_llm"):
+        self.text, self.error, self.success, self.model_id_used = text, error, success, model_id_used
+
+def initialize_text_llms():
+    global GOOGLE_API_KEY, HF_TOKEN, GEMINI_TEXT_CONFIGURED, HF_TEXT_CONFIGURED, hf_inference_text_client
+    print("INFO: llm_services.py - Initializing Text LLM clients...")
+    if GOOGLE_API_KEY and GOOGLE_API_KEY.strip():
+        try:
+            genai.configure(api_key=GOOGLE_API_KEY)
+            GEMINI_TEXT_CONFIGURED = True
+            print("SUCCESS: llm_services.py - Google Gemini API (for text) configured.")
+        except Exception as e:
+            print(f"ERROR: llm_services.py - Failed to configure Google Gemini API: {e}")
+            GEMINI_TEXT_CONFIGURED = False
+    else:
+        print("WARNING: llm_services.py - STORYVERSE_GOOGLE_API_KEY not found or empty.")
+        GEMINI_TEXT_CONFIGURED = False
+
+    if HF_TOKEN and HF_TOKEN.strip():
+        try:
+            hf_inference_text_client = InferenceClient(token=HF_TOKEN)
+            HF_TEXT_CONFIGURED = True
+            print("SUCCESS: llm_services.py - Hugging Face InferenceClient (for text) initialized.")
+        except Exception as e:
+            print(f"ERROR: llm_services.py - Failed to initialize HF InferenceClient: {e}")
+            HF_TEXT_CONFIGURED = False
+    else:
+        print("WARNING: llm_services.py - STORYVERSE_HF_TOKEN not found or empty.")
+        HF_TEXT_CONFIGURED = False
+    print(f"INFO: llm_services.py - Text LLM Init complete. Gemini Text: {GEMINI_TEXT_CONFIGURED}, HF Text: {HF_TEXT_CONFIGURED}")
+
+def is_gemini_text_ready(): return GEMINI_TEXT_CONFIGURED
+def is_hf_text_ready(): return HF_TEXT_CONFIGURED
+
+def generate_text_gemini(prompt: str, model_id: str = "gemini-1.5-flash-latest", system_prompt: str = None, temperature: float = 0.7, max_tokens: int = 512) -> LLMTextResponse:
+    if not is_gemini_text_ready():
+        return LLMTextResponse(error="Gemini text API not configured.", success=False, model_id_used=model_id)
+    try:
+        model = genai.GenerativeModel(model_name=model_id, system_instruction=system_prompt)
+        config = genai.types.GenerationConfig(temperature=temperature, max_output_tokens=max_tokens)
+        response = model.generate_content(prompt, generation_config=config)
+        # Add robust response checking as in AlgoForge's llm_clients.py
+        if response.prompt_feedback and response.prompt_feedback.block_reason:
+            return LLMTextResponse(error=f"Gemini: Prompt blocked ({response.prompt_feedback.block_reason})", success=False, model_id_used=model_id)
+        if not response.candidates or not response.candidates[0].content.parts:
+            return LLMTextResponse(error=f"Gemini: No content generated (Finish reason: {response.candidates[0].finish_reason if response.candidates else 'Unknown'})", success=False, model_id_used=model_id)
+        return LLMTextResponse(text=response.text, model_id_used=model_id)
+    except Exception as e:
+        return LLMTextResponse(error=f"Gemini API Error ({model_id}): {type(e).__name__} - {str(e)}", success=False, model_id_used=model_id)
+
+def generate_text_hf(prompt: str, model_id: str = "mistralai/Mistral-7B-Instruct-v0.2", system_prompt: str = None, temperature: float = 0.7, max_tokens: int = 512) -> LLMTextResponse:
+    if not is_hf_text_ready() or not hf_inference_text_client:
+        return LLMTextResponse(error="HF text API not configured.", success=False, model_id_used=model_id)
+    
+    full_prompt = f"<s>[INST] <<SYS>>\n{system_prompt}\n<</SYS>>\n\n{prompt} [/INST]" if system_prompt else prompt
+    try:
+        use_sample = temperature > 0.001
+        response_text = hf_inference_text_client.text_generation(
+            full_prompt, model=model_id, max_new_tokens=max_tokens,
+            temperature=temperature if use_sample else None, do_sample=use_sample
+        )
+        return LLMTextResponse(text=response_text, model_id_used=model_id)
+    except Exception as e:
+        return LLMTextResponse(error=f"HF API Error ({model_id}): {type(e).__name__} - {str(e)}", success=False, model_id_used=model_id)
+
+print("DEBUG: core.llm_services (for StoryVerseWeaver) - Module defined.")