rag-sentinel 0.1.2__tar.gz → 0.1.3__tar.gz

{rag_sentinel-0.1.2/src/rag_sentinel.egg-info → rag_sentinel-0.1.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rag-sentinel
-Version: 0.1.2
+Version: 0.1.3
 Summary: RAG Evaluation Framework using Ragas metrics and MLflow tracking
 Author: RAGSentinel Team
 License: MIT

{rag_sentinel-0.1.2 → rag_sentinel-0.1.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "rag-sentinel"
-version = "0.1.2"
+version = "0.1.3"
 description = "RAG Evaluation Framework using Ragas metrics and MLflow tracking"
 readme = "README.md"
 license = {text = "MIT"}

{rag_sentinel-0.1.2 → rag_sentinel-0.1.3}/src/rag_sentinel/__init__.py

@@ -20,5 +20,5 @@ Author: RAGSentinel Team
 License: MIT
 """
 
-__version__ = "0.1.0"
+__version__ = "0.1.2"
 

rag_sentinel-0.1.3/src/rag_sentinel/evaluator.py

@@ -0,0 +1,363 @@
+"""
+RAGSentinel Evaluator - Core evaluation logic.
+
+This module contains the main evaluation pipeline for RAGSentinel.
+It handles configuration loading, LLM initialization, API communication,
+Ragas metrics evaluation, and MLflow result logging.
+"""
+
+import os
+import re
+import yaml
+import configparser
+import requests
+import pandas as pd
+import mlflow
+from dotenv import load_dotenv
+from datasets import Dataset
+from ragas import evaluate
+from ragas.run_config import RunConfig
+from ragas.metrics import (
+    Faithfulness,
+    AnswerRelevancy,
+    ContextPrecision,
+    AnswerCorrectness,
+)
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings, AzureChatOpenAI, AzureOpenAIEmbeddings
+from langchain_ollama import ChatOllama, OllamaEmbeddings
+
+
+# =============================================================================
+# Configuration Loading
+# =============================================================================
+
+
+def load_config(yaml_file='rag_eval_config.yaml'):
+    """
+    Load configuration from YAML file with values resolved from .env and config.ini.
+
+    Returns:
+        dict: Fully resolved configuration dictionary
+    """
+    load_dotenv('.env')
+
+    ini = configparser.ConfigParser()
+    ini.read('config.ini')
+
+    def resolve(obj):
+        if isinstance(obj, dict):
+            return {k: resolve(v) for k, v in obj.items()}
+        if isinstance(obj, list):
+            return [resolve(i) for i in obj]
+        if isinstance(obj, str):
+            # Resolve ${ENV:VAR} and ${INI:section.key} placeholders
+            result = re.sub(r'\$\{ENV:([^}]+)\}', lambda m: os.getenv(m.group(1), ''), obj)
+            result = re.sub(r'\$\{INI:([^.]+)\.([^}]+)\}',
+                           lambda m: ini.get(m.group(1), m.group(2), fallback=''), result)
+            # Convert types
+            if result.lower() == 'true': return True
+            if result.lower() == 'false': return False
+            try:
+                if '.' in result: return float(result)
+            except ValueError:
+                pass
+            return result
+        return obj
+
+    with open(yaml_file, 'r') as f:
+        return resolve(yaml.safe_load(f))
+
+
+def get_llm(config):
+    """Initialize LLM based on config."""
+    provider = config['ragas']['llm']['provider']
+
+    if provider == "azure":
+        azure_config = config['ragas']['llm']['azure']
+        return AzureChatOpenAI(
+            azure_endpoint=azure_config['endpoint'],
+            api_key=azure_config['api_key'],
+            deployment_name=azure_config['deployment_name'],
+            model=azure_config['model'],
+            temperature=azure_config['temperature'],
+            api_version=azure_config['api_version']
+        )
+    elif provider == "openai":
+        openai_config = config['ragas']['llm']['openai']
+        return ChatOpenAI(
+            model=openai_config['model'],
+            temperature=openai_config['temperature'],
+            api_key=openai_config['api_key']
+        )
+    elif provider == "ollama":
+        ollama_config = config['ragas']['llm']['ollama']
+        return ChatOllama(
+            base_url=ollama_config['base_url'],
+            model=ollama_config['model'],
+            temperature=ollama_config['temperature']
+        )
+    else:
+        raise ValueError(f"Unsupported LLM provider: {provider}")
+
+
+def get_embeddings(config):
+    """Initialize embeddings based on config."""
+    provider = config['ragas']['embeddings']['provider']
+
+    if provider == "azure":
+        azure_config = config['ragas']['embeddings']['azure']
+        return AzureOpenAIEmbeddings(
+            azure_endpoint=azure_config['endpoint'],
+            api_key=azure_config['api_key'],
+            deployment=azure_config['deployment_name'],
+            api_version=azure_config['api_version']
+        )
+    elif provider == "openai":
+        openai_config = config['ragas']['embeddings']['openai']
+        return OpenAIEmbeddings(
+            model=openai_config['model'],
+            api_key=openai_config['api_key']
+        )
+    elif provider == "ollama":
+        ollama_config = config['ragas']['embeddings']['ollama']
+        return OllamaEmbeddings(
+            base_url=ollama_config['base_url'],
+            model=ollama_config['model']
+        )
+    else:
+        raise ValueError(f"Unsupported embeddings provider: {provider}")
+
+
+def get_metrics(config):
+    """Get Ragas metrics based on config."""
+    metric_map = {
+        "Faithfulness": Faithfulness(),
+        "AnswerRelevancy": AnswerRelevancy(),
+        "ContextPrecision": ContextPrecision(),
+        "AnswerCorrectness": AnswerCorrectness(),
+    }
+
+    metric_names = config['ragas']['metrics']
+    return [metric_map[name] for name in metric_names if name in metric_map]
+
+
+def get_auth_headers_and_cookies(config):
+    """Get authentication headers and cookies based on config."""
+    auth_config = config['backend']['auth']
+    auth_type = auth_config.get('type', 'none')
+
+    headers = {}
+    cookies = {}
+
+    if auth_type == "cookie":
+        cookies[auth_config['cookie_name']] = auth_config['cookie_value']
+    elif auth_type == "bearer":
+        headers['Authorization'] = f"Bearer {auth_config['bearer_token']}"
+    elif auth_type == "header":
+        headers[auth_config['header_name']] = auth_config['header_value']
+
+    return headers, cookies
+
+
+def extract_response_data(response, endpoint_config):
+    """Extract data from API response."""
+    if endpoint_config.get('stream', False):
+        return "".join(chunk.decode() for chunk in response.iter_content(chunk_size=None))
+
+    # Try to parse as JSON first
+    try:
+        data = response.json()
+        response_key = endpoint_config.get('response_key')
+        if response_key:
+            return data.get(response_key)
+        return data
+    except:
+        # If JSON parsing fails, return as plain text
+        return response.text
+
+
+def make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl=True):
+    """Make API request to backend."""
+    url = base_url + endpoint_config['path']
+    method = endpoint_config.get('method', 'POST')
+
+    headers = {**endpoint_config.get('headers', {}), **auth_headers}
+
+    # Flexible body preparation
+    body = {}
+    for key, value in endpoint_config.get('body', {}).items():
+        if isinstance(value, str) and ("{query}" in value or "{chat_id}" in value):
+            body[key] = value.format(query=query, chat_id=chat_id)
+        elif key == "chat_id":
+            try:
+                body[key] = int(chat_id)
+            except (ValueError, TypeError):
+                body[key] = chat_id
+        else:
+            body[key] = value
+
+    if method.upper() == 'POST':
+        resp = requests.post(
+            url,
+            json=body,
+            headers=headers,
+            cookies=auth_cookies,
+            stream=endpoint_config.get('stream', False),
+            verify=verify_ssl
+        )
+    elif method.upper() == 'GET':
+        resp = requests.get(
+            url,
+            params=body,
+            headers=headers,
+            cookies=auth_cookies,
+            verify=verify_ssl
+        )
+    else:
+        raise ValueError(f"Unsupported HTTP method: {method}")
+
+    resp.raise_for_status()
+    return resp
+
+
+def get_context(config, query, chat_id, auth_headers, auth_cookies):
+    """Retrieve context from backend API."""
+    base_url = config['backend']['base_url']
+    endpoint_config = config['backend']['endpoints']['context']
+    verify_ssl = config['backend'].get('verify_ssl', True)
+
+    response = make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl)
+    context = extract_response_data(response, endpoint_config)
+
+    if isinstance(context, str):
+        return [context]
+    elif isinstance(context, list):
+        return context
+    else:
+        return [str(context)]
+
+
+def get_answer(config, query, chat_id, auth_headers, auth_cookies):
+    """Get answer from backend API."""
+    base_url = config['backend']['base_url']
+    endpoint_config = config['backend']['endpoints']['answer']
+    verify_ssl = config['backend'].get('verify_ssl', True)
+
+    response = make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl)
+    answer = extract_response_data(response, endpoint_config)
+
+    return str(answer)
+
+
+
+def run_evaluation():
+    """Main evaluation function."""
+    print("=" * 60)
+    print("RAGSentinel - RAG Evaluation Framework")
+    print("=" * 60)
+
+    print("\n📁 Loading configuration...")
+    config = load_config()
+
+    dataset_path = config['dataset']['path']
+    print(f"📊 Loading dataset from {dataset_path}...")
+    dataset = pd.read_csv(dataset_path)
+
+    auth_headers, auth_cookies = get_auth_headers_and_cookies(config)
+
+    results = []
+    print(f"\n🔗 Collecting responses from {config['backend']['base_url']}...")
+
+    for idx, row in dataset.iterrows():
+        chat_id = str(row['chat_id'])
+        query = row['query']
+        ground_truth = row['ground_truth']
+
+        try:
+            context = get_context(config, query, chat_id, auth_headers, auth_cookies)
+            answer = get_answer(config, query, chat_id, auth_headers, auth_cookies)
+
+            results.append({
+                'question': query,
+                'contexts': context,
+                'answer': answer,
+                'ground_truth': ground_truth
+            })
+            print(f"  ✓ Processed query {idx + 1}/{len(dataset)}: {query[:50]}...")
+        except Exception as e:
+            print(f"  ✗ Error processing query {idx + 1}: {e}")
+            continue
+
+    if not results:
+        print("\n❌ No results collected. Exiting.")
+        return
+
+    eval_df = pd.DataFrame(results)
+    print(f"\n✓ Collected {len(eval_df)} responses")
+
+    print("\n🤖 Initializing LLM and embeddings...")
+    llm = get_llm(config)
+    embeddings = get_embeddings(config)
+
+    metrics = get_metrics(config)
+    print(f"   Metrics: {', '.join(config['ragas']['metrics'])}")
+
+    print("\n📈 Preparing data for RAGAS evaluation...")
+    ragas_data = {"question": [], "answer": [], "contexts": [], "ground_truth": []}
+
+    for _, row in eval_df.iterrows():
+        contexts = row.get("contexts", [])
+        if not isinstance(contexts, list):
+            contexts = [str(contexts)]
+        contexts = [str(c) for c in contexts if c and str(c).strip()]
+        if not contexts:
+            contexts = ["No context available."]
+
+        ragas_data["question"].append(str(row["question"]))
+        ragas_data["answer"].append(str(row["answer"]))
+        ragas_data["contexts"].append(contexts)
+        ragas_data["ground_truth"].append(str(row["ground_truth"]))
+
+    dataset = Dataset.from_dict(ragas_data)
+
+    print("\n⏳ Evaluating with Ragas metrics (this may take a while)...")
+
+    run_config = RunConfig(timeout=300, max_retries=3, max_wait=600)
+
+    ragas_result = evaluate(
+        dataset,
+        metrics=metrics,
+        llm=llm,
+        embeddings=embeddings,
+        batch_size=2,
+        run_config=run_config,
+        raise_exceptions=False
+    )
+
+    print("\n📊 Processing results...")
+    scores_df = ragas_result.to_pandas()
+    numeric_columns = scores_df.select_dtypes(include=['float64', 'float32', 'int64', 'int32']).columns
+    mean_scores = scores_df[numeric_columns].mean().to_dict()
+
+    mlflow_config = config['mlflow']
+    mlflow.set_tracking_uri(mlflow_config['tracking_uri'])
+    mlflow.set_experiment(mlflow_config['experiment_name'])
+
+    print("\n📤 Logging results to MLflow...")
+    run_name = mlflow_config.get('run_name', 'RAG Evaluation')
+    with mlflow.start_run(run_name=run_name):
+        print("\n" + "=" * 40)
+        print("📊 EVALUATION RESULTS")
+        print("=" * 40)
+        for metric_name, value in mean_scores.items():
+            mlflow.log_metric(metric_name, value)
+            print(f"   {metric_name}: {value:.4f}")
+
+        mlflow.log_param("dataset_path", dataset_path)
+        mlflow.log_param("num_samples", len(eval_df))
+        mlflow.log_table(data=scores_df, artifact_file="ragas_detailed_results.json")
+
+    print("\n" + "=" * 60)
+    print("✅ Evaluation complete!")
+    print(f"🔗 View results at: {mlflow_config['tracking_uri']}")
+    print("=" * 60)

{rag_sentinel-0.1.2 → rag_sentinel-0.1.3/src/rag_sentinel.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rag-sentinel
-Version: 0.1.2
+Version: 0.1.3
 Summary: RAG Evaluation Framework using Ragas metrics and MLflow tracking
 Author: RAGSentinel Team
 License: MIT

rag_sentinel-0.1.2/src/rag_sentinel/evaluator.py

@@ -1,392 +0,0 @@
-"""
-RAGSentinel Evaluator - Core evaluation logic.
-
-This module contains the main evaluation pipeline for RAGSentinel.
-It handles configuration loading, LLM initialization, API communication,
-Ragas metrics evaluation, and MLflow result logging.
-"""
-
-import os
-import re
-import yaml
-import configparser
-import requests
-import pandas as pd
-import mlflow
-from dotenv import load_dotenv
-from datasets import Dataset
-from ragas import evaluate, RunConfig
-from ragas.metrics import faithfulness, answer_relevancy, context_precision, answer_correctness
-
-
-# =============================================================================
-# Configuration Loading
-# =============================================================================
-
-
-def resolve_placeholder(value, env_vars, ini_config):
-    """
-    Resolve ${ENV:...} and ${INI:...} placeholders in a string value.
-
-    Args:
-        value: String that may contain placeholders
-        env_vars: Dictionary of environment variables
-        ini_config: ConfigParser object with ini file contents
-
-    Returns:
-        str: Value with all placeholders resolved
-    """
-    if not isinstance(value, str):
-        return value
-
-    # Resolve ${ENV:VAR_NAME} - reads from environment variables
-    env_pattern = r'\$\{ENV:([^}]+)\}'
-    def env_replacer(match):
-        var_name = match.group(1)
-        return env_vars.get(var_name, '')
-    value = re.sub(env_pattern, env_replacer, value)
-
-    # Resolve ${INI:section.key} - reads from config.ini
-    ini_pattern = r'\$\{INI:([^}]+)\}'
-    def ini_replacer(match):
-        path = match.group(1)
-        parts = path.split('.')
-        if len(parts) == 2:
-            section, key = parts
-            if ini_config.has_option(section, key):
-                return ini_config.get(section, key)
-        return ''
-    value = re.sub(ini_pattern, ini_replacer, value)
-
-    return value
-
-
-def resolve_config(obj, env_vars, ini_config):
-    """
-    Recursively resolve all placeholders in a configuration object.
-
-    Args:
-        obj: Configuration object (dict, list, or str)
-        env_vars: Dictionary of environment variables
-        ini_config: ConfigParser object
-
-    Returns:
-        Configuration object with all placeholders resolved
-    """
-    if isinstance(obj, dict):
-        return {k: resolve_config(v, env_vars, ini_config) for k, v in obj.items()}
-    elif isinstance(obj, list):
-        return [resolve_config(item, env_vars, ini_config) for item in obj]
-    elif isinstance(obj, str):
-        return resolve_placeholder(obj, env_vars, ini_config)
-    return obj
-
-
-def load_config():
-    """
-    Load and merge configuration from .env, config.ini, and rag_eval_config.yaml.
-
-    Returns:
-        dict: Fully resolved configuration dictionary
-    """
-    # Load environment variables from .env file
-    load_dotenv('.env')
-    env_vars = dict(os.environ)
-
-    # Load INI configuration
-    ini_config = configparser.ConfigParser()
-    ini_config.read('config.ini')
-
-    # Load YAML configuration and resolve all placeholders
-    with open('rag_eval_config.yaml', 'r') as f:
-        yaml_config = yaml.safe_load(f)
-
-    return resolve_config(yaml_config, env_vars, ini_config)
-
-
-def get_llm(config):
-    """Initialize LLM based on provider."""
-    llm_config = config['ragas']['llm']
-    provider = llm_config['provider'].lower()
-
-    if provider == 'azure':
-        from langchain_openai import AzureChatOpenAI
-        return AzureChatOpenAI(
-            azure_endpoint=llm_config['azure_endpoint'],
-            api_key=llm_config['api_key'],
-            api_version=llm_config.get('api_version', '2024-02-15-preview'),
-            deployment_name=llm_config['model'],
-            temperature=float(llm_config.get('temperature', 0.0))
-        )
-    elif provider == 'openai':
-        from langchain_openai import ChatOpenAI
-        return ChatOpenAI(
-            api_key=llm_config['api_key'],
-            model=llm_config['model'],
-            temperature=float(llm_config.get('temperature', 0.0))
-        )
-    elif provider == 'ollama':
-        from langchain_ollama import ChatOllama
-        return ChatOllama(
-            base_url=llm_config.get('base_url', 'http://localhost:11434'),
-            model=llm_config['model'],
-            temperature=float(llm_config.get('temperature', 0.0))
-        )
-    else:
-        raise ValueError(f"Unknown LLM provider: {provider}")
-
-
-def get_embeddings(config):
-    """Initialize embeddings based on provider."""
-    emb_config = config['ragas']['embeddings']
-    provider = emb_config['provider'].lower()
-
-    if provider == 'azure':
-        from langchain_openai import AzureOpenAIEmbeddings
-        return AzureOpenAIEmbeddings(
-            azure_endpoint=emb_config['azure_endpoint'],
-            api_key=emb_config['api_key'],
-            api_version=emb_config.get('api_version', '2024-02-15-preview'),
-            deployment=emb_config['model']
-        )
-    elif provider == 'openai':
-        from langchain_openai import OpenAIEmbeddings
-        return OpenAIEmbeddings(
-            api_key=emb_config['api_key'],
-            model=emb_config['model']
-        )
-    elif provider == 'ollama':
-        from langchain_ollama import OllamaEmbeddings
-        return OllamaEmbeddings(
-            base_url=emb_config.get('base_url', 'http://localhost:11434'),
-            model=emb_config['model']
-        )
-    else:
-        raise ValueError(f"Unknown embeddings provider: {provider}")
-
-
-def get_metrics(config):
-    """Get list of Ragas metrics."""
-    metric_map = {
-        'faithfulness': faithfulness,
-        'answer_relevancy': answer_relevancy,
-        'context_precision': context_precision,
-        'answer_correctness': answer_correctness
-    }
-    return [metric_map[m] for m in config['ragas']['metrics'] if m in metric_map]
-
-
-def get_auth_headers_and_cookies(config):
-    """
-    Get authentication headers and cookies from backend config.
-
-    Supports three authentication types:
-        - cookie: Session cookie authentication
-        - bearer: Bearer token authentication
-        - header: Custom header authentication
-
-    Args:
-        config: Full configuration dictionary
-
-    Returns:
-        tuple: (headers dict, cookies dict)
-    """
-    # Auth config is nested under backend.auth in the YAML
-    auth_config = config.get('backend', {}).get('auth', {})
-    auth_type = auth_config.get('type', 'none').lower()
-    headers = {}
-    cookies = {}
-
-    if auth_type == 'cookie':
-        cookie_name = auth_config.get('cookie_name', 'session')
-        cookie_value = auth_config.get('cookie_value', '')
-        if cookie_value:
-            cookies[cookie_name] = cookie_value
-    elif auth_type == 'bearer':
-        token = auth_config.get('bearer_token', '')
-        if token:
-            headers['Authorization'] = f'Bearer {token}'
-    elif auth_type == 'header':
-        header_name = auth_config.get('header_name', '')
-        header_value = auth_config.get('header_value', '')
-        if header_name and header_value:
-            headers[header_name] = header_value
-
-    return headers, cookies
-
-
-def extract_response_data(response, endpoint_config):
-    """Extract data from API response."""
-    data = response.json()
-    response_path = endpoint_config.get('response_path', '')
-
-    if response_path:
-        for key in response_path.split('.'):
-            if isinstance(data, dict) and key in data:
-                data = data[key]
-            elif isinstance(data, list) and key.isdigit():
-                data = data[int(key)]
-            else:
-                return data
-    return data
-
-
-def make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl=True):
-    """Make API request to backend."""
-    url = base_url.rstrip('/') + endpoint_config['path']
-    method = endpoint_config.get('method', 'POST').upper()
-
-    body = endpoint_config.get('body', {}).copy()
-    body['query'] = query
-    body['chat_id'] = chat_id
-
-    headers = {'Content-Type': 'application/json'}
-    headers.update(auth_headers)
-
-    if method == 'POST':
-        response = requests.post(url, json=body, headers=headers, cookies=auth_cookies, verify=verify_ssl)
-    else:
-        response = requests.get(url, params=body, headers=headers, cookies=auth_cookies, verify=verify_ssl)
-
-    response.raise_for_status()
-    return response
-
-
-def get_context(config, query, chat_id, auth_headers, auth_cookies):
-    """Get context from backend API."""
-    base_url = config['backend']['base_url']
-    endpoint_config = config['backend']['endpoints']['context']
-    verify_ssl = config['backend'].get('verify_ssl', True)
-
-    response = make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl)
-    context = extract_response_data(response, endpoint_config)
-
-    if isinstance(context, list):
-        return [str(c) for c in context]
-    return [str(context)]
-
-
-def get_answer(config, query, chat_id, auth_headers, auth_cookies):
-    """Get answer from backend API."""
-    base_url = config['backend']['base_url']
-    endpoint_config = config['backend']['endpoints']['answer']
-    verify_ssl = config['backend'].get('verify_ssl', True)
-
-    response = make_api_request(base_url, endpoint_config, query, chat_id, auth_headers, auth_cookies, verify_ssl)
-    answer = extract_response_data(response, endpoint_config)
-
-    return str(answer)
-
-
-
-def run_evaluation():
-    """Main evaluation function."""
-    print("=" * 60)
-    print("RAGSentinel - RAG Evaluation Framework")
-    print("=" * 60)
-
-    print("\n📁 Loading configuration...")
-    config = load_config()
-
-    dataset_path = config['dataset']['path']
-    print(f"📊 Loading dataset from {dataset_path}...")
-    dataset = pd.read_csv(dataset_path)
-
-    auth_headers, auth_cookies = get_auth_headers_and_cookies(config)
-
-    results = []
-    print(f"\n🔗 Collecting responses from {config['backend']['base_url']}...")
-
-    for idx, row in dataset.iterrows():
-        chat_id = str(row['chat_id'])
-        query = row['query']
-        ground_truth = row['ground_truth']
-
-        try:
-            context = get_context(config, query, chat_id, auth_headers, auth_cookies)
-            answer = get_answer(config, query, chat_id, auth_headers, auth_cookies)
-
-            results.append({
-                'question': query,
-                'contexts': context,
-                'answer': answer,
-                'ground_truth': ground_truth
-            })
-            print(f"  ✓ Processed query {idx + 1}/{len(dataset)}: {query[:50]}...")
-        except Exception as e:
-            print(f"  ✗ Error processing query {idx + 1}: {e}")
-            continue
-
-    if not results:
-        print("\n❌ No results collected. Exiting.")
-        return
-
-    eval_df = pd.DataFrame(results)
-    print(f"\n✓ Collected {len(eval_df)} responses")
-
-    print("\n🤖 Initializing LLM and embeddings...")
-    llm = get_llm(config)
-    embeddings = get_embeddings(config)
-
-    metrics = get_metrics(config)
-    print(f"   Metrics: {', '.join(config['ragas']['metrics'])}")
-
-    print("\n📈 Preparing data for RAGAS evaluation...")
-    ragas_data = {"question": [], "answer": [], "contexts": [], "ground_truth": []}
-
-    for _, row in eval_df.iterrows():
-        contexts = row.get("contexts", [])
-        if not isinstance(contexts, list):
-            contexts = [str(contexts)]
-        contexts = [str(c) for c in contexts if c and str(c).strip()]
-        if not contexts:
-            contexts = ["No context available."]
-
-        ragas_data["question"].append(str(row["question"]))
-        ragas_data["answer"].append(str(row["answer"]))
-        ragas_data["contexts"].append(contexts)
-        ragas_data["ground_truth"].append(str(row["ground_truth"]))
-
-    dataset = Dataset.from_dict(ragas_data)
-
-    print("\n⏳ Evaluating with Ragas metrics (this may take a while)...")
-
-    run_config = RunConfig(timeout=300, max_retries=3, max_wait=600)
-
-    ragas_result = evaluate(
-        dataset,
-        metrics=metrics,
-        llm=llm,
-        embeddings=embeddings,
-        batch_size=2,
-        run_config=run_config,
-        raise_exceptions=False
-    )
-
-    print("\n📊 Processing results...")
-    scores_df = ragas_result.to_pandas()
-    numeric_columns = scores_df.select_dtypes(include=['float64', 'float32', 'int64', 'int32']).columns
-    mean_scores = scores_df[numeric_columns].mean().to_dict()
-
-    mlflow_config = config['mlflow']
-    mlflow.set_tracking_uri(mlflow_config['tracking_uri'])
-    mlflow.set_experiment(mlflow_config['experiment_name'])
-
-    print("\n📤 Logging results to MLflow...")
-    run_name = mlflow_config.get('run_name', 'RAG Evaluation')
-    with mlflow.start_run(run_name=run_name):
-        print("\n" + "=" * 40)
-        print("📊 EVALUATION RESULTS")
-        print("=" * 40)
-        for metric_name, value in mean_scores.items():
-            mlflow.log_metric(metric_name, value)
-            print(f"   {metric_name}: {value:.4f}")
-
-        mlflow.log_param("dataset_path", dataset_path)
-        mlflow.log_param("num_samples", len(eval_df))
-        mlflow.log_table(data=scores_df, artifact_file="ragas_detailed_results.json")
-
-    print("\n" + "=" * 60)
-    print("✅ Evaluation complete!")
-    print(f"🔗 View results at: {mlflow_config['tracking_uri']}")
-    print("=" * 60)