lollms-client 0.33.0__py3-none-any.whl → 1.1.0__py3-none-any.whl

examples/internet_search_with_rag.py

@@ -1,226 +0,0 @@
-from lollms_client import LollmsClient, MSG_TYPE
-from ascii_colors import ASCIIColors, trace_exception
-from typing import List, Dict, Any, Optional, Callable
-import json
-from pathlib import Path
-
-# --- Internet Search RAG Implementation ---
-_duckduckgo_search_installed = False
-_search_installation_error_message = ""
-try:
-    import pipmaster as pm
-    # ensure_packages should be called by the binding init ideally,
-    # but we call it here for the example's standalone execution.
-    pm.ensure_packages(["duckduckgo_search"]) 
-    from duckduckgo_search import DDGS
-    _duckduckgo_search_installed = True
-except Exception as e:
-    _search_installation_error_message = str(e)
-    DDGS = None
-    ASCIIColors.error(f"Failed to import duckduckgo_search: {_search_installation_error_message}")
-    ASCIIColors.info("Please install it: pip install duckduckgo-search")
-
-
-def perform_internet_search_rag(
-    query_text: str,
-    vectorizer_name: Optional[str] = None, # Not used for search
-    top_k: int = 5,
-    min_similarity_percent: float = 0.0 # Not used directly for search filter, but can influence result quality/rank
-) -> List[Dict[str, Any]]:
-    """
-    Performs an internet search using DuckDuckGo and formats results for RAG.
-    Similarity is simulated based on rank.
-    """
-    if not _duckduckgo_search_installed or DDGS is None:
-         ASCIIColors.error("duckduckgo_search is not available. Cannot perform internet search.")
-         return []
-
-    ASCIIColors.magenta(f"  [INTERNET SEARCH] Querying DuckDuckGo for: '{query_text}', max_results={top_k}")
-    search_results_raw = []
-    try:
-        # DDGS().text returns a generator, max_results limits it.
-        # Note: The DDGS library might sometimes return fewer results than max_results.
-        with DDGS() as ddgs:
-            search_results_raw = list(ddgs.text(keywords=query_text, max_results=top_k))
-            
-    except Exception as e:
-        ASCIIColors.error(f"  [INTERNET SEARCH] Search failed: {e}")
-        trace_exception(e)
-        return []
-
-    formatted_results: List[Dict[str, Any]] = []
-    if search_results_raw:
-        for i, r in enumerate(search_results_raw):
-            # Simulate similarity based on rank (rank 1 is highest sim)
-            # Max similarity is 100% for rank 1, decreases linearly or non-linearly.
-            # Simple linear decrease: 100 - (rank * (100 / top_k+1))
-            # Let's use rank-based score: 100% for rank 1, 90% for rank 2, ... 50% for rank 5 etc.
-            # Ensure similarity is above min_similarity_percent if that param was intended as a filter here
-            
-            simulated_similarity = max(0.0, 100.0 - i * (100.0 / (top_k + 1))) # Higher rank = lower sim
-            simulated_similarity = round(simulated_similarity, 2)
-
-            if simulated_similarity >= min_similarity_percent:
-                formatted_results.append({
-                    "file_path": r.get("href", "# Unknown URL"), # Use URL as document identifier
-                    "chunk_text": f"Title: {r.get('title', 'N/A')}\nSnippet: {r.get('body', 'N/A')}", # Combine title and snippet
-                    "similarity_percent": simulated_similarity,
-                })
-            else:
-                 ASCIIColors.debug(f"  [INTERNET SEARCH] Skipping result {i+1} due to low simulated similarity ({simulated_similarity}%)")
-
-    ASCIIColors.magenta(f"  [INTERNET SEARCH] Formatted {len(formatted_results)} results for RAG.")
-    if not formatted_results: ASCIIColors.yellow(f"  [INTERNET SEARCH] No results found for query: '{query_text}' or none met min_similarity_percent.")
-    return formatted_results
-
-# --- Streaming Callback for RAG and LLM ---
-def rag_streaming_callback(
-    chunk: str, 
-    msg_type: MSG_TYPE, 
-    metadata: Optional[Dict] = None, 
-    turn_history: Optional[List] = None
-) -> bool:
-    metadata = metadata or {}
-    hop = metadata.get("hop", "")
-    type_info = metadata.get("type", "N/A")
-
-    if msg_type == MSG_TYPE.MSG_TYPE_CHUNK: # Final answer chunks
-        ASCIIColors.success(chunk, end="", flush=True)
-    elif msg_type == MSG_TYPE.MSG_TYPE_STEP_START:
-        info = metadata.get("query", chunk) if type_info in ["rag_query_generation", "rag_retrieval"] else chunk
-        ASCIIColors.yellow(f"\n>> RAG Hop {hop} | START | {type_info.upper()} | Info: {str(info)[:100]}...", flush=True)
-    elif msg_type == MSG_TYPE.MSG_TYPE_STEP_END:
-        num_chunks = metadata.get("num_chunks")
-        query = metadata.get("query")
-        decision = metadata.get("decision")
-        
-        end_info = []
-        if query: end_info.append(f"Query: '{str(query)[:50]}...'")
-        if num_chunks is not None: end_info.append(f"Results: {num_chunks}")
-        if decision: end_info.append(f"LLM Decision: NeedMore={decision.get('need_more_data')}, Summary: '{str(decision.get('new_information_summary'))[:50]}...'")
-        
-        ASCIIColors.green(f"\n<< RAG Hop {hop} | END   | {type_info.upper()} | {' | '.join(end_info) if end_info else chunk}", flush=True)
-    elif msg_type == MSG_TYPE.MSG_TYPE_EXCEPTION:
-        ASCIIColors.error(f"\nError in RAG stream: {chunk}", flush=True)
-    
-    return True
-
-# --- Main Example ---
-if __name__ == "__main__":
-    ASCIIColors.red("--- Multi-Hop Internet Search Example with LollmsClient ---")
-
-    # LLM Configuration (use a model good at instruction following and JSON)
-    # Ensure your Ollama server is running and has this model pulled.
-    LLM_BINDING_NAME = "ollama"
-    LLM_MODEL_NAME = "mistral:latest" # or llama3, phi3 etc.
-
-    # You could also enable the internet_search tool via MCP,
-    # but this example specifically uses it directly via generate_text_with_rag.
-    # For MCP example, see examples/local_mcp.py
-
-    try:
-        lc = LollmsClient(
-            binding_name=LLM_BINDING_NAME,
-            model_name=LLM_MODEL_NAME,
-            temperature=0.1,
-            ctx_size=4096 
-        )
-        ASCIIColors.green(f"LollmsClient initialized with LLM: {LLM_BINDING_NAME}/{LLM_MODEL_NAME}")
-
-        if not _duckduckgo_search_installed or DDGS is None:
-             ASCIIColors.error("duckduckgo_search is not installed. Cannot run search examples.")
-             exit()
-
-
-        # --- Test Case 1: Classic Search RAG (max_rag_hops = 0) ---
-        ASCIIColors.cyan("\n\n--- Test Case 1: Classic Internet Search RAG (max_rag_hops = 0) ---")
-        classic_search_prompt = "What is the current population of Japan?"
-        ASCIIColors.blue(f"User Prompt: {classic_search_prompt}")
-
-        classic_rag_result = lc.generate_text_with_rag(
-            prompt=classic_search_prompt,
-            rag_query_function=perform_internet_search_rag, # Use the search function
-            max_rag_hops=0,
-            rag_top_k=3, # Get 3 search results
-            rag_min_similarity_percent=50.0, # Only use results with simulated sim >= 50%
-            streaming_callback=rag_streaming_callback,
-            n_predict=250
-        )
-        print("\n--- End of Classic Search RAG ---")
-        ASCIIColors.magenta("\nClassic Search RAG Final Output Structure:")
-        print(f"  Final Answer (first 100 chars): {classic_rag_result.get('final_answer', '')}...")
-        print(f"  Error: {classic_rag_result.get('error')}")
-        print(f"  Number of Hops: {len(classic_rag_result.get('rag_hops_history', []))}")
-        print(f"  Total Unique Sources Retrieved: {len(classic_rag_result.get('all_retrieved_sources', []))}")
-        if classic_rag_result.get('all_retrieved_sources'):
-            print("  Example Retrieved Source:")
-            source_ex = classic_rag_result['all_retrieved_sources'][0]
-            print(f"    Document (URL): {source_ex.get('document')}")
-            print(f"    Similarity: {source_ex.get('similarity')}%")
-            print(f"    Content (Snippet, first 50 chars): {source_ex.get('content', '')}...")
-
-
-        # --- Test Case 2: Multi-Hop Search RAG (max_rag_hops = 1) ---
-        ASCIIColors.cyan("\n\n--- Test Case 2: Multi-Hop Internet Search RAG (max_rag_hops = 1) ---")
-        multihop_search_prompt_1 = "Tell me about the latest developments in fusion energy, including any recent news."
-        ASCIIColors.blue(f"User Prompt: {multihop_search_prompt_1}")
-
-        multihop_rag_result_1 = lc.generate_text_with_rag(
-            prompt=multihop_search_prompt_1,
-            rag_query_function=perform_internet_search_rag,
-            rag_query_text=None, # LLM will generate first query
-            max_rag_hops=1, # Allow one refinement hop
-            rag_top_k=2, # Get 2 search results per query
-            rag_min_similarity_percent=50.0,
-            streaming_callback=rag_streaming_callback,
-            n_predict=400
-        )
-        print("\n--- End of Multi-Hop Search RAG (1 hop max) ---")
-        ASCIIColors.magenta("\nMulti-Hop Search RAG (1 hop max) Final Output Structure:")
-        print(f"  Final Answer (first 100 chars): {multihop_rag_result_1.get('final_answer', '')}...")
-        print(f"  Error: {multihop_rag_result_1.get('error')}")
-        print(f"  Number of Hops Made: {len(multihop_rag_result_1.get('rag_hops_history', []))}")
-        for i, hop_info in enumerate(multihop_rag_result_1.get('rag_hops_history', [])):
-            print(f"    Hop {i+1} Query: '{hop_info.get('query')}'")
-            print(f"    Hop {i+1} Results Count: {len(hop_info.get('retrieved_chunks_details',[]))}")
-            print(f"    Hop {i+1} Summary (first 50): '{str(hop_info.get('new_information_summary'))[:50]}...'")
-            print(f"    Hop {i+1} LLM Decision: NeedMoreData={hop_info.get('llm_decision_json',{}).get('need_more_data')}")
-        print(f"  Total Unique Sources Retrieved: {len(multihop_rag_result_1.get('all_retrieved_sources', []))}")
-
-
-        # --- Test Case 3: More complex multi-hop (max_rag_hops = 2) ---
-        ASCIIColors.cyan("\n\n--- Test Case 3: More Complex Multi-Hop Internet Search RAG (max_rag_hops = 2) ---")
-        multihop_search_prompt_2 = "What are the requirements and steps to install the lollms_client python library, and what are some of its key features?"
-        ASCIIColors.blue(f"User Prompt: {multihop_search_prompt_2}")
-
-        multihop_rag_result_2 = lc.generate_text_with_rag(
-            prompt=multihop_search_prompt_2,
-            rag_query_function=perform_internet_search_rag,
-            max_rag_hops=2, # Allow up to two refinement hops
-            rag_top_k=2, # Get 2 results per query
-            rag_min_similarity_percent=40.0, # Lower similarity to maybe get broader initial results
-            streaming_callback=rag_streaming_callback,
-            n_predict=500 # Allow more for the installation steps and features
-        )
-        print("\n--- End of More Complex Multi-Hop Search RAG (up to 2 hops) ---")
-        ASCIIColors.magenta("\nMore Complex Multi-Hop Search RAG (up to 2 hops) Final Output Structure:")
-        print(f"  Final Answer (first 100 chars): {multihop_rag_result_2.get('final_answer', '')[:100]}...")
-        print(f"  Error: {multihop_rag_result_2.get('error')}")
-        print(f"  Number of Hops Made: {len(multihop_rag_result_2.get('rag_hops_history', []))}")
-        for i, hop_info in enumerate(multihop_rag_result_2.get('rag_hops_history', [])):
-            print(f"    Hop {i+1} Query: '{hop_info.get('query')}'")
-            print(f"    Hop {i+1} Results Count: {len(hop_info.get('retrieved_chunks_details',[]))}")
-            print(f"    Hop {i+1} Summary (first 50): '{str(hop_info.get('new_information_summary'))[:50]}...'")
-        print(f"  Total Unique Sources Retrieved: {len(multihop_rag_result_2.get('all_retrieved_sources', []))}")
-
-
-    except ValueError as ve:
-        ASCIIColors.error(f"Initialization or RAG parameter error: {ve}")
-        trace_exception(ve)
-    except ConnectionRefusedError:
-        ASCIIColors.error(f"Connection refused. Is the Ollama server ({LLM_BINDING_NAME}) running?")
-    except Exception as e:
-        ASCIIColors.error(f"An unexpected error occurred: {e}")
-        trace_exception(e)
-
-    ASCIIColors.red("\n--- Multi-Hop Internet Search Example Finished ---")

examples/lollms_chat/calculator.py

@@ -1,59 +0,0 @@
-# This is a simple example of a calculator using lollms_chat. don't expect this to be perfect as LLMs are very bad in computing 
-# this can be used for very simple calculations. don't expect it to be accurate to compute sqrt(35) or something
-# it is just for the purpose of experimentation as even the best models in the world are not capable of doing accurate calculations yet without a calculator
-
-from lollms_client import LollmsClient
-import json
-import math  # Import the math module for calculations
-
-# Make sure you use your key
-lc = LollmsClient(
-    "openai",
-    "http://localhost:9642/v1/",
-    service_key="lollms_y-uyV-p2_AQGo5Ut6uHDmfIoRk6rKfmf0Rz6xQx-Zkl8cNyVUSFM"# make sure you generate your own key
-)
-
-# if you want to see what binding/model does the server support, use this:
-models = lc.listModels()
-print(f"Found models:\n{models}")
-
-lc.set_model_name("ollama/gemma3:27b")  # Or your preferred binding/model
-
-expression = input("Give an expression to evaluate: ")
-
-# Construct a detailed prompt
-system_prompt = (
-    "You are a highly accurate calculator.  You receive a mathematical expression "
-    "as input and return the result as a JSON object.  "
-    "The expression can include numbers, basic arithmetic operators (+, -, *, /), "
-    "parentheses, and common mathematical functions like sin, cos, tan, pi, sqrt, and log.  "
-    "Always evaluate the expression and return the final numeric result.  If the expression is invalid, return 'Error'."
-)
-
-template = '{"result": the numeric result of the evaluated expression}'
-
-# Include the expression in the user prompt.  This is important!
-user_prompt = f"Evaluate the following expression: {expression}"
-
-# Generate the code
-generation_output = lc.generate_code(
-    user_prompt,
-    system_prompt=system_prompt,
-    template=template
-)
-
-try:
-    # Attempt to parse the JSON response
-    generation_output = json.loads(generation_output)
-    result = generation_output["result"]
-
-    # Attempt to convert the result to a float
-    try:
-        result = float(result)
-        print(f"Result: {result}")
-    except ValueError:
-        print(f"Result: {result} (Could not convert to a number)") #Handles cases where the LLM returns non-numeric output
-except json.JSONDecodeError:
-    print(f"Error: Could not decode JSON response: {generation_output}")
-except KeyError:
-    print(f"Error: 'result' key not found in JSON response: {generation_output}")

examples/lollms_chat/derivative.py

@@ -1,48 +0,0 @@
-from lollms_client import LollmsClient
-import json
-
-# Make sure you use your key
-lc = LollmsClient(
-    "openai",
-    "http://localhost:9642/v1/",
-    service_key="lollms_y-uyV-p2_AQGo5Ut6uHDmfIoRk6rKfmf0Rz6xQx-Zkl8cNyVUSFM"  # make sure you generate your own key
-)
-
-# if you want to see what binding/model does the server support, use this:
-models = lc.listModels()
-print(f"Found models:\n{models}")
-
-lc.set_model_name("ollama/gemma3:27b")  # Or your preferred binding/model
-
-function = input("Enter the function (e.g., x^2 + 2*x): ")
-parameter = input("Enter the parameter to differentiate with respect to (e.g., x): ")
-
-# Construct a detailed prompt
-system_prompt = (
-    "You are a symbolic differentiation engine. You receive a mathematical function "
-    "and a parameter as input, and you return the derivative of the function with respect to that parameter. "
-    "The function can include variables, numbers, and common mathematical operations. "
-    "Return the derivative as a string. If the function or parameter is invalid, return 'Error'."
-)
-
-template = '"{derivative}": the derivative of the function with respect to the parameter'
-
-# Include the function and parameter in the user prompt. This is important!
-user_prompt = f"Find the derivative of the function '{function}' with respect to '{parameter}'."
-
-# Generate the code
-generation_output = lc.generate_code(
-    user_prompt,
-    system_prompt=system_prompt,
-    template=template
-)
-
-try:
-    # Attempt to parse the JSON response
-    generation_output = json.loads(generation_output)
-    derivative = generation_output["derivative"]
-    print(f"Derivative: {derivative}")
-except json.JSONDecodeError:
-    print(f"Error: Could not decode JSON response: {generation_output}")
-except KeyError:
-    print(f"Error: 'derivative' key not found in JSON response: {generation_output}")

examples/lollms_chat/test_openai_compatible_with_lollms_chat.py

@@ -1,12 +0,0 @@
-from lollms_client import LollmsClient
-#make sure you use your key
-lc = LollmsClient("openai","http://localhost:9642/v1/", service_key="lollms_zXQdyvrP_ecMXm3UZ0D004x979aHpyF8iq4ki_b52q0WdFuiEfMo")
-models = lc.listModels()
-print(f"Found models:\n{models}")
-
-lc.set_model_name("ollama/gemma3:27b")
-
-res = lc.generate_text("Describe this image",images=[
-    r"C:\Users\parisneo\Pictures\me.jpg"
-])
-print(res)

examples/lollms_discussions_test.py

@@ -1,155 +0,0 @@
-# research_app_final.py
-
-import os
-import json
-import shutil
-from pathlib import Path
-# Use the correct, specified import style
-from lollms_client import LollmsClient
-from lollms_client.lollms_discussion import LollmsDataManager, LollmsDiscussion
-from lollms_client.lollms_types import MSG_TYPE
-from sqlalchemy import Column, String
-
-# --- 1. Define Application-Specific Schema ---
-# The developer can define their own fields for the database tables.
-# This allows applications to store and query their own metadata.
-class ResearchDiscussionMixin:
-    # We want each discussion to have a 'project_name' that we can search for.
-    project_name = Column(String(100), index=True, nullable=False)
-
-class ResearchMessageMixin:
-    # This mixin is empty for this example.
-    pass
-
-def setup_migration_dummies(folder: Path):
-    """Creates a dummy JSON file to simulate an old, file-based project structure."""
-    if not folder.exists():
-        folder.mkdir(parents=True, exist_ok=True)
-    
-    # This data structure mimics what the old `to_dict` would have produced.
-    discussion_data = {
-        "id": "old_project_alpha",
-        "metadata": {"project_name": "Project Alpha"},
-        "system_prompt": "This is the system prompt for Alpha.",
-        "created_at": "2023-01-01T12:00:00",
-        "updated_at": "2023-01-01T12:05:00",
-        "messages": [
-            {"id": "msg1", "sender": "user", "sender_type":"user", "content": "What was the first finding?", "created_at": "2023-01-01T12:00:00"},
-            {"id": "msg2", "sender": "assistant", "sender_type":"assistant", "content": "It was about quantum states.", "parent_id": "msg1", "created_at": "2023-01-01T12:05:00"}
-        ]
-    }
-    with open(folder / "project_alpha.json", "w") as f:
-        json.dump(discussion_data, f, indent=2)
-    print(f"Created dummy migration file in '{folder}'.")
-
-def main():
-    # --- 2. Setup: Lollms Client is always needed ---
-    print("--- LOLLMS Research Assistant (Final Version) ---")
-    try:
-        # Instantiate the real LollmsClient to connect to a running model service.
-        # Ensure Ollama is running and has pulled the specified model.
-        lc = LollmsClient("ollama", model_name="mistral-nemo:latest")
-        print("LollmsClient connected successfully to Ollama.")
-    except Exception as e:
-        print(f"\nFATAL: Could not connect to LLM binding. Is Ollama running?\nError: {e}")
-        return
-
-    # --- DEMO 1: In-Memory Mode (Backward Compatibility) ---
-    print("\n--- DEMO 1: In-Memory Discussion ---")
-    
-    # Create an in-memory discussion by NOT passing a db_manager.
-    in_memory_discussion = LollmsDiscussion.create_new(lollms_client=lc)
-    in_memory_discussion.system_prompt = "You are a helpful assistant."
-    print("Created an in-memory discussion.")
-
-    # Interact with it. The state is held entirely in the object.
-    user_input_mem = "Can you remember that my favorite color is blue?"
-    print(f"You > {user_input_mem}")
-    print("AI  > ", end="", flush=True)
-    def stream_to_console(token, msg_type=MSG_TYPE.MSG_TYPE_CHUNK):
-        print(token, end="", flush=True)
-        return True
-    in_memory_discussion.chat(user_input_mem, streaming_callback=stream_to_console)
-    print()
-
-    # Save its state to a JSON file. This now works correctly.
-    file_path = Path("./in_memory_save.json")
-    with open(file_path, "w") as f:
-        json.dump(in_memory_discussion.to_dict(), f, indent=2)
-    print(f"\nIn-memory discussion saved to '{file_path}'.")
-    os.remove(file_path)
-
-    # --- DEMO 2: Database-Backed Mode with Migration ---
-    print("\n--- DEMO 2: Database-Backed Mode ---")
-    DB_PATH = "sqlite:///research_projects_final.db"
-    ENCRYPTION_KEY = "a-secure-password-for-the-database"
-    MIGRATION_FOLDER = Path("./old_discussions")
-    
-    try:
-        # Initialize the LollmsDataManager with our schema and encryption key.
-        db_manager = LollmsDataManager(
-            db_path=DB_PATH,
-            discussion_mixin=ResearchDiscussionMixin,
-            message_mixin=ResearchMessageMixin,
-            encryption_key=ENCRYPTION_KEY
-        )
-        print(f"Database setup complete. Encryption is ENABLED.")
-    except Exception as e:
-        print(f"\nFATAL: Could not initialize database. Error: {e}")
-        return
-
-    # Demonstrate the one-time migration from a folder of JSON files.
-    setup_migration_dummies(MIGRATION_FOLDER)
-    input("\nDummy migration files created. Press Enter to run the migration...")
-    LollmsDiscussion.migrate(lollms_client=lc, db_manager=db_manager, folder_path=MIGRATION_FOLDER)
-    
-    session = db_manager.get_session()
-    migrated_count = session.query(db_manager.DiscussionModel).count()
-    print(f"Verification: Found {migrated_count} discussions in the database after migration.")
-    session.close()
-
-    # --- DEMO 3: Live Chat with a DB-Backed Discussion ---
-    input("\nMigration complete. Press Enter to start a new, database-backed chat session...")
-    
-    # Create a new, database-backed discussion with our custom 'project_name'.
-    discussion = LollmsDiscussion.create_new(
-        lollms_client=lc,
-        db_manager=db_manager,
-        max_context_size=lc.default_ctx_size // 2,
-        autosave=True,
-        project_name="Project Gamma (Live)"
-    )
-    discussion.system_prompt = "You are a helpful assistant for Project Gamma."
-    
-    print(f"\n--- Live Chat for '{discussion.db_discussion.project_name}' ---")
-    print("Type your message, or '/exit', '/export_openai', '/export_ollama' to quit.")
-    
-    while True:
-        user_input = input("\nYou > ")
-        if user_input.lower() == '/exit': break
-        
-        if user_input.lower().startswith('/export'):
-            try:
-                format_type = user_input.split('_')[1] + "_chat"
-                exported_data = discussion.export(format_type)
-                print(f"\n--- Exported for {format_type.split('_')[0].upper()} ---")
-                print(json.dumps(exported_data, indent=2))
-                print("-----------------------------------")
-            except IndexError:
-                print("Invalid export command. Use /export_openai or /export_ollama")
-            continue
-
-        print("AI > ", end="", flush=True)
-        # The same streaming callback works seamlessly.
-        discussion.chat(user_input, streaming_callback=stream_to_console)
-        print()
-
-    # --- Cleanup ---
-    print("\n--- Demo complete. Cleaning up. ---")
-    if os.path.exists(DB_PATH):
-        os.remove(DB_PATH)
-    if MIGRATION_FOLDER.exists():
-        shutil.rmtree(MIGRATION_FOLDER)
-
-if __name__ == "__main__":
-    main()