llumo 0.2.14b7__py3-none-any.whl → 0.2.15__py3-none-any.whl

llumo/client.py

@@ -5,12 +5,12 @@ import time
 import re
 import json
 import uuid
-
+import warnings
 import os
 import itertools
 import pandas as pd
 from typing import List, Dict
-from .models import AVAILABLEMODELS, getProviderFromModel
+from .models import AVAILABLEMODELS, getProviderFromModel, Provider
 from .execution import ModelExecutor
 from .exceptions import LlumoAIError
 from .helpingFuntions import *
@@ -19,6 +19,8 @@ from .functionCalling import LlumoAgentExecutor
 import threading
 from tqdm import tqdm
 
+pd.set_option('future.no_silent_downcasting', True)
+
 postUrl = (
     "https://red-skull-service-392377961931.us-central1.run.app/api/process-playground"
 )
@@ -38,7 +40,7 @@ class LlumoClient:
 
     def __init__(self, api_key):
         self.apiKey = api_key
-        self.socket = LlumoSocketClient(socketUrl)
+        
         self.processMapping = {}
         self.definationMapping = {}
 
@@ -50,6 +52,7 @@ class LlumoClient:
         reqBody = {"analytics": [evalName]}
 
         try:
+            
             response = requests.post(url=validateUrl, json=reqBody, headers=headers)
 
         except requests.exceptions.RequestException as e:
@@ -581,7 +584,8 @@ class LlumoClient:
         createExperiment: bool = False,
         _tocheck=True,
     ):
-        dataframe = pd.DataFrame(data)
+        self.socket = LlumoSocketClient(socketUrl)
+        dataframe = pd.DataFrame(data).astype(str)
         workspaceID = None
         email = None
         socketID = self.socket.connect(timeout=250)
@@ -774,9 +778,6 @@ class LlumoClient:
             rawResults.extend(dataFromDb)
             
             
-            
-            
-            
 
         # Initialize dataframe columns for each eval
         for eval in evals:
@@ -797,7 +798,12 @@ class LlumoClient:
 
         if createExperiment:
             pd.set_option("future.no_silent_downcasting", True)
-            df = dataframe.fillna("Some error occured").astype(object)
+            # df = dataframe.fillna("Some error occured").astype(object)
+            with warnings.catch_warnings():
+                warnings.simplefilter(action='ignore', category=FutureWarning)
+                df = dataframe.fillna("Some error occurred").astype(str)
+
+            df = dataframe.fillna("Some error occured").infer_objects(copy=False)
             if createPlayground(
                 email,
                 workspaceID,
@@ -812,7 +818,7 @@ class LlumoClient:
         else:
             return dataframe
 
-    def run_sweep(
+    def promptSweep(
         self,
         templates: List[str],
         dataset: Dict[str, List[str]],
@@ -821,9 +827,15 @@ class LlumoClient:
         evals=["Response Correctness"],
         toEvaluate: bool = False,
         createExperiment: bool = False,
+
+
     ) -> pd.DataFrame:
 
-        self.validateApiKey(evalName=" ")
+        modelStatus = validateModels(model_aliases=model_aliases)
+        if modelStatus["status"]== False:
+            raise LlumoAIError.providerError(modelStatus["message"])
+
+        self.validateApiKey()
         workspaceID = self.workspaceID
         email = self.email
         executor = ModelExecutor(apiKey)
@@ -928,6 +940,7 @@ class LlumoClient:
         evals=["Final Task Alignment"],
         prompt_template="Give answer for the given query: {{query}}",
         createExperiment: bool = False,
+
     ):
         if model.lower() not in ["openai", "google"]:
             raise ValueError("Model must be 'openai' or 'google'")
@@ -1002,174 +1015,367 @@ class LlumoClient:
         except Exception as e:
             raise e
 
-    def runDataStream(
-        self,
-        data,
-        streamName: str,
-        queryColName: str = "query",
-        createExperiment: bool = False,
+    def ragSweep(
+            self,
+            data,
+            streamName: str,
+            queryColName: str = "query",
+            createExperiment: bool = False,
+            modelAliases=[],
+            apiKey="",
+            prompt_template="Give answer to the given: {{query}} using the context:{{context}}",
+            evals=["Context Utilization"],
+            toEvaluate=False,
+            generateOutput=True
     ):
-        results = {}
-        dataframe = pd.DataFrame(data)
-        try:
-            socketID = self.socket.connect(timeout=150)
-            # Ensure full connection before proceeding
-            max_wait_secs = 20
-            waited_secs = 0
-            while not self.socket._connection_established.is_set():
-                time.sleep(0.1)
-                waited_secs += 0.1
-                if waited_secs >= max_wait_secs:
-                    raise RuntimeError(
-                        "Timeout waiting for server 'connection-established' event."
-                    )
-            # print(f"Connected with socket ID: {socketID}")
-            rowIdMapping = {}
+        # Validate required parameters
+        if generateOutput:
+            if not modelAliases:
+                raise ValueError("Model aliases must be provided when generateOutput is True.")
+            if not apiKey or not isinstance(apiKey, str) or apiKey.strip() == "":
+                raise ValueError("Valid API key must be provided when generateOutput is True.")
+
+        modelStatus = validateModels(model_aliases=modelAliases)
+        if modelStatus["status"]== False:
+            if len(modelAliases) ==  0:
+                raise LlumoAIError.providerError("No model selected.")
+            else:
+                raise LlumoAIError.providerError(modelStatus["message"])
 
-            # print(f"Validating API key...")
-            self.validateApiKey()
-            # print(f"API key validation successful. Hits available: {self.hitsAvailable}")
+        # Copy the original dataframe
+        original_df = pd.DataFrame(data)
+        working_df = original_df.copy()
 
-            # check for available hits and trial limit
-            userHits = checkUserHits(
-                self.workspaceID,
-                self.hasSubscribed,
-                self.trialEndDate,
-                self.subscriptionEndDate,
-                self.hitsAvailable,
-                len(dataframe),
-            )
+        # Connect to socket
+        self.socket = LlumoSocketClient(socketUrl)
+        socketID = self.socket.connect(timeout=150)
+        waited_secs = 0
+        while not self.socket._connection_established.is_set():
+            time.sleep(0.1)
+            waited_secs += 0.1
+            if waited_secs >= 20:
+                raise RuntimeError("Timeout waiting for server 'connection-established' event.")
 
-            # do not proceed if subscription or trial limit has exhausted
-            if not userHits["success"]:
-                raise LlumoAIError.InsufficientCredits(userHits["message"])
+        self.validateApiKey()
 
-            print("====🚀Sit back while we fetch data from the stream 🚀====")
-            workspaceID = self.workspaceID
-            email = self.email
-            streamId = getStreamId(workspaceID, self.apiKey, streamName)
-            # Prepare all batches before sending
-            # print("Preparing batches...")
-            self.allBatches = []
-            currentBatch = []
+        # Check user credits
+        userHits = checkUserHits(
+            self.workspaceID, self.hasSubscribed, self.trialEndDate,
+            self.subscriptionEndDate, self.hitsAvailable, len(working_df)
+        )
+        if not userHits["success"]:
+            raise LlumoAIError.InsufficientCredits(userHits["message"])
 
-            for index, row in dataframe.iterrows():
-                activePlayground = f"{int(time.time() * 1000)}{uuid.uuid4()}".replace(
-                    "-", ""
-                )
-                rowID = f"{int(time.time() * 1000)}{uuid.uuid4()}".replace("-", "")
-                columnID = f"{int(time.time() * 1000)}{uuid.uuid4()}".replace("-", "")
+        print("====🚀Sit back while we fetch data from the stream 🚀====")
+        workspaceID, email = self.workspaceID, self.email
+        activePlayground = f"{int(time.time() * 1000)}{uuid.uuid4()}".replace("-", "")
+        streamId = getStreamId(workspaceID, self.apiKey, streamName)
 
-                rowIdMapping[rowID] = index
-                # Use the server-provided socket ID here
-                templateData = {
-                    "processID": getProcessID(),
-                    "socketID": socketID,
-                    "processData": {
-                        "executionDependency": {"query": row[queryColName]},
-                        "dataStreamID": streamId,
-                    },
-                    "workspaceID": workspaceID,
-                    "email": email,
-                    "type": "DATA_STREAM",
-                    "playgroundID": activePlayground,
-                    "processType": "DATA_STREAM",
-                    "rowID": rowID,
-                    "columnID": columnID,
-                    "source": "SDK",
-                }
+        # Prepare batches
+        rowIdMapping = {}
+        self.allBatches = []
+        currentBatch = []
 
-                currentBatch.append(templateData)
+        expectedResults = len(working_df)
+        timeout = max(100, min(150, expectedResults * 10))
 
-            if len(currentBatch) == 10 or index == len(dataframe) - 1:
+        listener_thread = threading.Thread(
+            target=self.socket.listenForResults,
+            kwargs={
+                "min_wait": 40,
+                "max_wait": timeout,
+                "inactivity_timeout": 10,
+                "expected_results": expectedResults,
+            },
+            daemon=True
+        )
+        listener_thread.start()
+
+        for index, row in working_df.iterrows():
+            rowID, columnID = uuid.uuid4().hex, uuid.uuid4().hex
+            compoundKey = f"{rowID}-{columnID}-{columnID}"
+            rowIdMapping[compoundKey] = {"index": index}
+            templateData = {
+                "processID": getProcessID(),
+                "socketID": socketID,
+                "processData": {
+                    "executionDependency": {"query": row[queryColName]},
+                    "dataStreamID": streamId,
+                },
+                "workspaceID": workspaceID,
+                "email": email,
+                "type": "DATA_STREAM",
+                "playgroundID": activePlayground,
+                "processType": "DATA_STREAM",
+                "rowID": rowID,
+                "columnID": columnID,
+                "source": "SDK",
+            }
+            currentBatch.append(templateData)
+            if len(currentBatch) == 10 or index == len(working_df) - 1:
                 self.allBatches.append(currentBatch)
                 currentBatch = []
 
-            # Post all batches
-            total_items = sum(len(batch) for batch in self.allBatches)
-            # print(f"Posting {len(self.allBatches)} batches ({total_items} items total)")
+        for batch in tqdm(self.allBatches, desc="Processing Batches", unit="batch", colour="magenta", ncols=80):
+            try:
+                self.postDataStream(batch=batch, workspaceID=workspaceID)
+                time.sleep(3)
+            except Exception as e:
+                print(f"Error posting batch: {e}")
+                raise
 
-            for cnt, batch in enumerate(self.allBatches):
-                # print(f"Posting batch {cnt + 1}/{len(self.allBatches)} for eval '{eval}'")
+        time.sleep(3)
+        listener_thread.join()
+
+        rawResults = self.socket.getReceivedData()
+        expectedRowIDs = set(rowIdMapping.keys())
+        receivedRowIDs = {key for item in rawResults for key in item.keys()}
+        missingRowIDs = list(expectedRowIDs - receivedRowIDs)
+
+        if missingRowIDs:
+            dataFromDb = fetchData(workspaceID, activePlayground, missingRowIDs)
+            rawResults.extend(dataFromDb)
+
+        working_df["context"] = None
+        for item in rawResults:
+            for compound_key, value in item.items():
+                if compound_key in rowIdMapping:
+                    idx = rowIdMapping[compound_key]["index"]
+                    working_df.at[idx, "context"] = value.get("value")
+
+        # Output generation
+        if generateOutput == True:
+            working_df = self._outputForStream(working_df, modelAliases, prompt_template, apiKey)
+
+        # Optional evaluation
+        outputEvalMapping = None
+        if toEvaluate:
+            for evalName in evals:
+                # Validate API and dependencies
+                self.validateApiKey(evalName=evalName)
+                metricDependencies = checkDependency(
+                    evalName, list(working_df.columns), tocheck=False
+                )
+                if not metricDependencies["status"]:
+                    raise LlumoAIError.dependencyError(metricDependencies["message"])
+
+            working_df, outputEvalMapping = self._evaluateForStream(working_df, evals, modelAliases, prompt_template,generateOutput)
+
+        
+        self.socket.disconnect()
+
+        # Create experiment if required
+        if createExperiment:
+            # df = working_df.fillna("Some error occured").astype(object)
+            with warnings.catch_warnings():
+                warnings.simplefilter(action='ignore', category=FutureWarning)
+                df = working_df.fillna("Some error occurred").astype(str)
+            if createPlayground(
+                    email, workspaceID, df,
+                    queryColName=queryColName,
+                    dataStreamName=streamId,
+                    promptText=prompt_template,
+                    definationMapping=self.definationMapping,
+                    evalOutputMap=outputEvalMapping
+            ):
+                print(
+                    "Your data has been saved in the Llumo Experiment. Visit https://app.llumo.ai/evallm to see the results.")
+        else:
+            self.latestDataframe = working_df
+            return working_df
+
+    def _outputForStream(self, df, modelAliases, prompt_template, apiKey):
+        executor = ModelExecutor(apiKey)
+
+        for indx, row in df.iterrows():
+            inputVariables = re.findall(r"{{(.*?)}}", prompt_template)
+            if not all([k in df.columns for k in inputVariables]):
+                raise LlumoAIError.InvalidPromptTemplate()
+
+            inputDict = {key: row[key] for key in inputVariables}
+            for i, model in enumerate(modelAliases, 1):
                 try:
-                    self.postDataStream(batch=batch, workspaceID=workspaceID)
-                    # print(f"Batch {cnt + 1} posted successfully")
+
+                    provider = getProviderFromModel(model)
+                    if provider == Provider.OPENAI:
+                        validateOpenaiKey(apiKey)
+                    elif provider == Provider.GOOGLE:
+                        validateGoogleKey(apiKey)
+
+                    filled_template = getInputPopulatedPrompt(prompt_template, inputDict)
+                    response = executor.execute(provider, model.value, filled_template, apiKey)
+                    df.at[indx, f"output_{i}"] = response
+                    
                 except Exception as e:
-                    print(f"Error posting batch {cnt + 1}: {str(e)}")
-                    continue
+                    # df.at[indx, f"output_{i}"] = str(e)
+                    raise e
 
-                # Small delay between batches to prevent overwhelming the server
-                time.sleep(1)
+        return df
 
-            # updating the dict for row column mapping
-            self.AllProcessMapping()
-            # Calculate a reasonable timeout based on the data size
-            timeout = max(60, min(600, total_items * 10))
-            # print(f"All batches posted. Waiting up to {timeout} seconds for results...")
+    def _evaluateForStream(self, df, evals, modelAliases, prompt_template,generateOutput):
+        dfWithEvals = df.copy()
 
-            # Listen for results
-            self.socket.listenForResults(
-                min_wait=20,
-                max_wait=timeout,
-                inactivity_timeout=30,
-                expected_results=None,
-            )
+        outputColMapping = {}
+        for i, model in enumerate(modelAliases, 1):
+            if generateOutput:
+                outputColName = f"output_{i}" 
+            else:
+                outputColName = "output"
+            try:
 
-            # Get results for this evaluation
-            eval_results = self.socket.getReceivedData()
-            # print(f"Received {len(eval_results)} results for evaluation '{eval}'")
+                res = self.evaluateMultiple(
+                    dfWithEvals.to_dict("records"),
+                    evals=evals,
+                    prompt_template=prompt_template,
+                    outputColName=outputColName,
+                    _tocheck=False,
+                )
+                for evalMetric in evals:
+                    scoreCol = f"{evalMetric}"
+                    reasonCol = f"{evalMetric} Reason"
 
-            # Add these results to our overall results
-            results["Data Stream"] = self.finalResp(eval_results)
-            print(f"=======You are all set! continue your expectations 🚀======\n")
+                    if scoreCol in res.columns:
+                        res = res.rename(columns={scoreCol: f"{scoreCol}_{i}"})
+                    if reasonCol in res.columns:
+                        res = res.rename(columns={reasonCol: f"{evalMetric}_{i} Reason"})
 
-            # print("All evaluations completed successfully")
+                    outputColMapping[f"{scoreCol}_{i}"] = outputColName
 
-        except Exception as e:
-            print(f"Error during evaluation: {e}")
-            raise
-        finally:
-            # Always disconnect the socket when done
+                newCols = [col for col in res.columns if col not in dfWithEvals.columns]
+                dfWithEvals = pd.concat([dfWithEvals, res[newCols]], axis=1)
+            except Exception as e:
+                print(f"Evaluation failed for model {model.value}: {str(e)}")
+        return dfWithEvals, outputColMapping
+
+    def runDataStream(
+            self,
+            data,
+            streamName: str,
+            queryColName: str = "query",
+            createExperiment: bool = False,
+           ):
+
+
+        # Copy the original dataframe
+        original_df = pd.DataFrame(data)
+        working_df = original_df.copy()
+
+        # Connect to socket
+        self.socket = LlumoSocketClient(socketUrl)
+        socketID = self.socket.connect(timeout=150)
+        waited_secs = 0
+        while not self.socket._connection_established.is_set():
+            time.sleep(0.1)
+            waited_secs += 0.1
+            if waited_secs >= 20:
+                raise RuntimeError("Timeout waiting for server 'connection-established' event.")
+
+        self.validateApiKey()
+
+        # Check user credits
+        userHits = checkUserHits(
+            self.workspaceID, self.hasSubscribed, self.trialEndDate,
+            self.subscriptionEndDate, self.hitsAvailable, len(working_df)
+        )
+        if not userHits["success"]:
+            raise LlumoAIError.InsufficientCredits(userHits["message"])
+
+        print("====🚀Sit back while we fetch data from the stream 🚀====")
+        workspaceID, email = self.workspaceID, self.email
+        activePlayground = f"{int(time.time() * 1000)}{uuid.uuid4()}".replace("-", "")
+        streamId = getStreamId(workspaceID, self.apiKey, streamName)
+
+        # Prepare batches
+        rowIdMapping = {}
+        self.allBatches = []
+        currentBatch = []
+
+        expectedResults = len(working_df)
+        timeout = max(100, min(150, expectedResults * 10))
+
+        listener_thread = threading.Thread(
+            target=self.socket.listenForResults,
+            kwargs={
+                "min_wait": 40,
+                "max_wait": timeout,
+                "inactivity_timeout": 10,
+                "expected_results": expectedResults,
+            },
+            daemon=True
+        )
+        listener_thread.start()
+
+        for index, row in working_df.iterrows():
+            rowID, columnID = uuid.uuid4().hex, uuid.uuid4().hex
+            compoundKey = f"{rowID}-{columnID}-{columnID}"
+            rowIdMapping[compoundKey] = {"index": index}
+            templateData = {
+                "processID": getProcessID(),
+                "socketID": socketID,
+                "processData": {
+                    "executionDependency": {"query": row[queryColName]},
+                    "dataStreamID": streamId,
+                },
+                "workspaceID": workspaceID,
+                "email": email,
+                "type": "DATA_STREAM",
+                "playgroundID": activePlayground,
+                "processType": "DATA_STREAM",
+                "rowID": rowID,
+                "columnID": columnID,
+                "source": "SDK",
+            }
+            currentBatch.append(templateData)
+            if len(currentBatch) == 10 or index == len(working_df) - 1:
+                self.allBatches.append(currentBatch)
+                currentBatch = []
+
+        for batch in tqdm(self.allBatches, desc="Processing Batches", unit="batch", colour="magenta", ncols=80):
             try:
-                self.socket.disconnect()
-                # print("Socket disconnected")
+                self.postDataStream(batch=batch, workspaceID=workspaceID)
+                time.sleep(3)
             except Exception as e:
-                print(f"Error disconnecting socket: {e}")
+                print(f"Error posting batch: {e}")
+                raise
 
-        for streamName, records in results.items():
-            dataframe[streamName] = None
-            for item in records:
-                for compound_key, value in item.items():
-                    # for compound_key, value in item['data'].items():
+        time.sleep(3)
+        listener_thread.join()
 
-                    rowID = compound_key.split("-")[0]
-                    # looking for the index of each rowID , in the original dataframe
-                    if rowID in rowIdMapping:
-                        index = rowIdMapping[rowID]
-                        # dataframe.at[index, evalName] = value
-                        dataframe.at[index, streamName] = value["value"]
+        rawResults = self.socket.getReceivedData()
+        expectedRowIDs = set(rowIdMapping.keys())
+        receivedRowIDs = {key for item in rawResults for key in item.keys()}
+        missingRowIDs = list(expectedRowIDs - receivedRowIDs)
 
-                    else:
-                        pass
-                        # print(f"⚠️ Warning: Could not find rowID {rowID} in mapping")
+        if missingRowIDs:
+            dataFromDb = fetchData(workspaceID, activePlayground, missingRowIDs)
+            rawResults.extend(dataFromDb)
 
-        if createExperiment:
-            pd.set_option("future.no_silent_downcasting", True)
-            df = dataframe.fillna("Some error occured").astype(object)
+        working_df["context"] = None
+        for item in rawResults:
+            for compound_key, value in item.items():
+                if compound_key in rowIdMapping:
+                    idx = rowIdMapping[compound_key]["index"]
+                    working_df.at[idx, "context"] = value.get("value")
 
+
+
+        self.socket.disconnect()
+
+        # Create experiment if required
+        if createExperiment:
+            df = working_df.fillna("Some error occured").astype(object)
             if createPlayground(
-                email,
-                workspaceID,
-                df,
-                queryColName=queryColName,
-                dataStreamName=streamId,
+                    email, workspaceID, df,
+                    queryColName=queryColName,
+                    dataStreamName=streamId,
+                    definationMapping=self.definationMapping,
             ):
                 print(
-                    "Your data has been saved in the Llumo Experiment. Visit https://app.llumo.ai/evallm to see the results."
-                )
+                    "Your data has been saved in the Llumo Experiment. Visit https://app.llumo.ai/evallm to see the results.")
         else:
-            self.latestDataframe = dataframe
-            return dataframe
+            self.latestDataframe = working_df
+            return working_df
+
 
     def createExperiment(self, dataframe):
         try:

llumo/exceptions.py

@@ -50,6 +50,10 @@ class LlumoAIError(Exception):
     def dependencyError(details):
         return LlumoAIError(details)
 
+    @staticmethod
+    def providerError(details):
+        return LlumoAIError(details)
+
     # @staticmethod
     # def dateNotFound():
     #     return LlumoAIError("Trial end date or subscription end date not found for the given user.")

llumo/execution.py

@@ -25,15 +25,14 @@ class ModelExecutor:
         return response.choices[0].message.content
 
     def _executeGoogle(self, modelName: str, prompt: str,api_key) -> str:
-        
+
         # Configure GenAI with API Key
         genai.configure(api_key=api_key)
-            
+
         # Select Generative Model
         model = genai.GenerativeModel("gemini-2.0-flash-lite")
         # Generate Response
         response = model.generate_content(prompt)
         return response.text
-        
-    
- 
+
+

llumo/helpingFuntions.py

@@ -8,7 +8,11 @@ import json
 import base64
 import os
 import re
+import openai
+import google.generativeai as genai
 
+
+from .models import  _MODEL_METADATA,  AVAILABLEMODELS
 subscriptionUrl = "https://app.llumo.ai/api/workspace/record-extra-usage"
 getStreamdataUrl = "https://app.llumo.ai/api/data-stream/all"
 createPlayUrl = "https://app.llumo.ai/api/New-Eval-API/create-new-eval-playground"
@@ -212,7 +216,8 @@ def deleteColumnListInPlayground(workspaceID: str, playgroundID: str):
         print("❌ Error:", response.status_code, response.text)
         return None
 
-def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColName=None,outputColName= "output",dataStreamName=None,definationMapping=None):
+def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColName=None,
+                 outputColName= "output",dataStreamName=None,definationMapping=None,evalOutputMap = None):
     if len(dataframe) > 100:
         dataframe = dataframe.head(100)
         print("⚠️ Dataframe truncated to 100 rows for upload.")
@@ -232,11 +237,11 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
     # Iterate over each column in the dataframe
     for indx, col in enumerate(dataframe.columns):
         # Generate a unique column ID using uuid
-        columnID = str(uuid.uuid4().hex[:8]) 
+        columnID = str(uuid.uuid4().hex[:8])
 
         columnIDMapping[col] = columnID
 
-        
+
         if col.startswith('output') and promptText!=None:
             # For output columns, create the prompt template with promptText
             if promptText:
@@ -248,12 +253,12 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
 
                 # Loop through each variable and check if it exists as a column name
                 for var in variables:
-                    varName = var.strip() 
+                    varName = var.strip()
                     if varName in columnIDMapping:  # Check if the variable is a column name
                         dependencies.append(columnIDMapping[varName])  # Add its columnID
 
                 # Now update the template for the output column
-                
+
                 template={
                     "provider": "OPENAI",
                     "model": "GPT_4o",
@@ -275,8 +280,8 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
                     "type": "PROMPT",
                     "order": indx,
                 }
-                
-        elif col.startswith('Data ') :
+
+        elif col.startswith('context') and dataStreamName != None :
             if queryColName and dataStreamName:
                 dependencies = []
                 dependencies.append(columnIDMapping[queryColName])
@@ -286,22 +291,27 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
                     "dataStreamName": dataStreamName,
                     "query": columnIDMapping[queryColName],
                     "columnID": columnID,  # Use the generated column ID
-                    "label": "Data stream",
+                    "label": "context",
                     "type": "DATA_STREAM",
                     "order": indx}
 
-        elif col in allEvals and promptText!=None:
 
+        elif any(col.startswith(eval + "_") or col == eval for eval in allEvals) and not " Reason" in col and promptText is not None:
+            if evalOutputMap != None:
+                outputColName = evalOutputMap[col]
+            else:
+                outputColName = outputColName
             dependencies = []
             variables = re.findall(r'{{(.*?)}}', promptText)
 
                 # Loop through each variable and check if it exists as a column name
             for var in variables:
-                varName = var.strip() 
+                varName = var.strip()
                 if varName in columnIDMapping:  # Check if the variable is a column name
                     dependencies.append(columnIDMapping[varName])
-                    
+
             dependencies.append(columnIDMapping[outputColName])  # Add the output column ID
+
             longDef = definationMapping.get(col, {}).get('definition', "")
             shortDef =definationMapping.get(col, {}).get('briefDefinition', "")
             enum =  col.upper().replace(" ","_")
@@ -341,11 +351,11 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
     }
 
         elif col.endswith(' Reason') and promptText!=None:
-            continue   
+            continue
+
 
-            
         else:
-            
+
             template = {
                 "label": col,  # Label is the column name
                 "type": "VARIABLE",  # Default type for non-output columns
@@ -370,25 +380,27 @@ def createColumn(workspaceID, dataframe, playgroundID, promptText=None,queryColN
         row_dict = {}
 
         # For each column, we need to map the column ID to the corresponding value in the row
+        
         for col in dataframe.columns:
             columnID = columnIDMapping[col]
-            
-            if col in allEvals and promptText!=None:
+
+            if any(col.startswith(eval + "_") or col == eval for eval in allEvals) and not " Reason"  in col and promptText!=None:
+                
                 row_dict[columnID] = {
-                    
+
                     "value": row[col],
                     "type": "EVAL",
                     "isValid": True,
                     "reasoning": row[col+" Reason"],
                     "edgeCase": "minorHallucinationDetailNotInContext",
                     "kpi": col
-                
-                    }
+
+                }
             elif col.endswith(' Reason') and promptText!=None:
                 continue
             else:# Get the columnID from the mapping
                 row_dict[columnID] = row[col]
-            
+
             # row_dict[columnID] = row[col]  # Directly map the column ID to the row value
         # Add the row index (if necessary)
         row_dict["pIndex"] = indx
@@ -440,11 +452,11 @@ def uploadRowsInDBPlayground(payload):
         return None
 
 
-def createPlayground(email, workspaceID, df, promptText=None,queryColName=None,dataStreamName=None,definationMapping=None,outputColName="output"):
+def createPlayground(email, workspaceID, df, promptText=None,queryColName=None,dataStreamName=None,definationMapping=None,outputColName="output",evalOutputMap = None):
     
     playgroundId = str(createEvalPlayground(email=email, workspaceID=workspaceID))
     payload1, payload2 = createColumn(
-        workspaceID=workspaceID, dataframe=df, playgroundID=playgroundId, promptText=promptText,queryColName=queryColName,dataStreamName=dataStreamName,definationMapping=definationMapping,outputColName=outputColName
+        workspaceID=workspaceID, dataframe=df, playgroundID=playgroundId, promptText=promptText,queryColName=queryColName,dataStreamName=dataStreamName,definationMapping=definationMapping,outputColName=outputColName,evalOutputMap=evalOutputMap
     )
     
  # Debugging line to check the payload2 structure
@@ -606,3 +618,36 @@ def fetchData(workspaceID, playgroundID, missingList: list):
     except Exception as e:
         print(f"An error occurred: {e}")
         return []
+
+def validateModels(model_aliases):
+
+    selectedProviders = []
+    for name in model_aliases:
+        for alias ,(provider , modelName ) in _MODEL_METADATA.items():
+            if modelName == name:
+                selectedProviders.append(provider)
+
+    if len(set(selectedProviders)) > 1:
+        return {"status": False,"message":"All selected models should be of same provider."}
+    else:
+        return {"status": True,"message":"All selected models are of same provider."}
+
+
+
+def validateOpenaiKey(api_key):
+    try:
+        client = openai.OpenAI(api_key=api_key)
+        _ = client.models.list()  # Light call to list models
+    except openai.AuthenticationError:
+        raise ValueError("❌ Invalid OpenAI API key.")
+    except Exception as e:
+        raise RuntimeError(f"⚠️ Error validating OpenAI key: {e}")
+
+def validateGoogleKey(api_key):
+    try:
+        genai.configure(api_key=api_key)
+        _ = genai.GenerativeModel("gemini-2.0-flash-lite").generate_content("test")
+    except Exception as e:
+        if "PERMISSION_DENIED" in str(e) or "API key not valid" in str(e):
+            raise ValueError("❌ Invalid Google API key.")
+        raise RuntimeError(f"⚠️ Error validating Gemini key: {e}")

llumo/models.py

@@ -6,35 +6,72 @@ class Provider(str, Enum):
 
 # Maps model aliases → (provider, actual model name for API)
 _MODEL_METADATA = {
-    "GPT_4": (Provider.OPENAI, "gpt-4"),
-    "GPT_4_32K": (Provider.OPENAI, "gpt-4-32k"),
-    "GPT_35T": (Provider.OPENAI, "gpt-3.5-turbo"),
-    "GPT_35T_INS": (Provider.OPENAI, "gpt-3.5-turbo-instruct"),
-    "GPT_35T_16K": (Provider.OPENAI, "gpt-3.5-turbo-16k"),
-    "GPT_35_TURBO": (Provider.OPENAI, "gpt-3.5-turbo"),
-
-    "GOOGLE_15_FLASH": (Provider.GOOGLE, "gemini-1.5-flash-latest"),
-    "GEMINI_PRO": (Provider.GOOGLE, "gemini-pro"),
-    "TEXT_BISON": (Provider.GOOGLE, "text-bison-001"),
-    "CHAT_BISON": (Provider.GOOGLE, "chat-bison-001"),
-    "TEXT_BISON_32K": (Provider.GOOGLE, "text-bison-32k"),
-    "TEXT_UNICORN": (Provider.GOOGLE, "text-unicorn-experimental"),
+    "GPT_4O": (Provider.OPENAI, "GPT_4O"),
+    "GPT_4_5": (Provider.OPENAI, "GPT_4_5"),
+    "GPT_4": (Provider.OPENAI, "GPT_4"),
+    "GPT_4_32K": (Provider.OPENAI, "GPT_4_32K"),
+    "GPT_3_5_Turbo": (Provider.OPENAI, "GPT_35T"),
+    "GPT_3_5_Turbo_Instruct": (Provider.OPENAI, "GPT_35T_INS"),
+    "GPT_3_5_Turbo_16K": (Provider.OPENAI, "GPT_35T_16K"),
+    "GPT_4_o_Mini": (Provider.OPENAI, "GPT_4O_MINI"),
+    "o4_MINI": (Provider.OPENAI, "O4_MINI"),
+    "o4_MINI_HIGH": (Provider.OPENAI, "O4_MINI_HIGH"),
+    "GPT_4_1": (Provider.OPENAI, "GPT_4_1"),
+    "GPT_4_1_Mini": (Provider.OPENAI, "GPT_4_1_MINI"),
+    "GPT_4_1_nano": (Provider.OPENAI, "GPT_4_1_NANO"),
+    "o3": (Provider.OPENAI, "O3"),
+    "o3_MINI": (Provider.OPENAI, "O3_MINI"),
+    "o1": (Provider.OPENAI, "O1"),
+    "o1_MINI": (Provider.OPENAI, "O1_MINI"),
+    
+    
+    "Gemini_2_5_Pro": (Provider.GOOGLE, "GEMINI_2_5_PRO"),
+    "Gemini_2_5_Flash": (Provider.GOOGLE, "GEMINI_2_5_FLASH"),
+    "Gemini_2_0": (Provider.GOOGLE, "GEMINI_2_0"),
+    "Gemini_2_0_Flash": (Provider.GOOGLE, "GEMINI_2_0_FLASH"),
+    "Gemini_Pro": (Provider.GOOGLE, "GEMINI_PRO"),
+    "Text_Bison": (Provider.GOOGLE, "TEXT_BISON"),
+    "Chat_Bison": (Provider.GOOGLE, "CHAT_BISON"),
+    "Text_Bison_32k": (Provider.GOOGLE, "TEXT_BISON_32K"),
+    "Text_Unicorn": (Provider.GOOGLE, "TEXT_UNICORN"),
+    "Google_1_5_Flash": (Provider.GOOGLE, "GOOGLE_15_FLASH"),
+    "Gemma_3_9B": (Provider.GOOGLE, "GEMMA_3_9B"),
+    "Gemma_3_27B": (Provider.GOOGLE, "GEMMA_3_27B"),
 }
 
 class AVAILABLEMODELS(str, Enum):
-    GPT_4 = "gpt-4"
-    GPT_4_32K = "gpt-4-32k"
-    GPT_35T = "gpt-3.5-turbo"
-    GPT_35T_INS = "gpt-3.5-turbo-instruct"
-    GPT_35T_16K = "gpt-3.5-turbo-16k"
-    GPT_35_TURBO = "gpt-3.5-turbo"
-
-    GOOGLE_15_FLASH = "gemini-1.5-flash-latest"
-    GEMINI_PRO = ""
-    TEXT_BISON = "text-bison-001"
-    CHAT_BISON = "chat-bison-001"
-    TEXT_BISON_32K = "text-bison-32k"
-    TEXT_UNICORN = "text-unicorn-experimental"
+    GPT_4o= "GPT_4O",
+    GPT_4o_Mini= "GPT_4O_MINI",
+    GPT_4_5= "GPT_4_5",
+    GPT_4= "GPT_4",
+    GPT_4_32K= "GPT_4_32K",
+    GPT_3_5_Turbo= "GPT_35T",
+    GPT_3_5_Turbo_Instruct= "GPT_35T_INS",
+    GPT_3_5_Turbo_16K= "GPT_35T_16K",
+    GPT_4_o_Mini= "GPT_4O_MINI",
+    o4_MINI = "O4_MINI",
+    o4_MINI_HIGH = "O4_MINI_HIGH",
+    GPT_4_1 = "GPT_4_1",
+    GPT_4_1_Mini = "GPT_4_1_MINI",
+    GPT_4_1_nano = "GPT_4_1_NANO",
+    o3 = "O3",
+    o3_MINI = "O3_MINI",
+    o1 = "O1",
+    o1_MINI = "O1_MINI",
+    
+    Gemini_2_5_Pro = "GEMINI_2_5_PRO",
+    Gemini_2_5_Flash = "GEMINI_2_5_FLASH",
+    Gemini_2_0 = "GEMINI_2_0",
+    Gemini_2_0_Flash = "GEMINI_2_0_FLASH",
+    Gemini_Pro = "GEMINI_PRO",
+    Text_Bison = "TEXT_BISON",
+    Chat_Bison = "CHAT_BISON",
+    Text_Bison_32k = "TEXT_BISON_32K",
+    Text_Unicorn = "TEXT_UNICORN",
+    Google_1_5_Flash = "GOOGLE_15_FLASH",
+    Gemma_3_9B = "GEMMA_3_9B",
+    Gemma_3_27B = "GEMMA_3_27B",
+    
 
 def getProviderFromModel(model: AVAILABLEMODELS) -> Provider:
     for alias, (provider, apiName) in _MODEL_METADATA.items():

{llumo-0.2.14b7.dist-info → llumo-0.2.15.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llumo
-Version: 0.2.14b7
+Version: 0.2.15
 Summary: Python SDK for interacting with the Llumo ai API.
 Home-page: https://www.llumo.ai/
 Author: Llumo

llumo-0.2.15.dist-info/RECORD

@@ -0,0 +1,13 @@
+llumo/__init__.py,sha256=O04b4yW1BnOvcHzxWFddAKhtdBEhBNhLdb6xgnpHH_Q,205
+llumo/client.py,sha256=zh6fpKpjlYcvzrPZkPviF1hDRzfnA1K0U1gweoKfkwc,54675
+llumo/exceptions.py,sha256=Vp_MnanHbnd1Yjuoi6WLrKiwwZbJL3znCox2URMmGU4,2032
+llumo/execution.py,sha256=nWbJ7AvWuUPcOb6i-JzKRna_PvF-ewZTiK8skS-5n3w,1380
+llumo/functionCalling.py,sha256=D5jYapu1rIvdIJNUYPYMTyhQ1H-6nkwoOLMi6eekfUE,7241
+llumo/helpingFuntions.py,sha256=-9GA9X0KBUVZb3_25D8AlninWnVc9ajFp4QkR_mDePY,23545
+llumo/models.py,sha256=aVEZsOOoQx5LeNtwSyBxqvrINq0izH3QWu_YjsMPE6o,2910
+llumo/sockets.py,sha256=I2JO_eNEctRo_ikgvFVp5zDd-m0VDu04IEUhhsa1Tic,5950
+llumo-0.2.15.dist-info/licenses/LICENSE,sha256=tF9yAcfPV9xGT3ViWmC8hPvOo8BEk4ZICbUfcEo8Dlk,182
+llumo-0.2.15.dist-info/METADATA,sha256=OQApH-0Gj918OaMbyQasOtE6lAhU5__No3SK9xge-NM,1519
+llumo-0.2.15.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+llumo-0.2.15.dist-info/top_level.txt,sha256=d5zUTMI99llPtLRB8rtSrqELm_bOqX-bNC5IcwlDk88,6
+llumo-0.2.15.dist-info/RECORD,,

llumo-0.2.14b7.dist-info/RECORD

@@ -1,13 +0,0 @@
-llumo/__init__.py,sha256=O04b4yW1BnOvcHzxWFddAKhtdBEhBNhLdb6xgnpHH_Q,205
-llumo/client.py,sha256=HpvUyucrGPbcPQMz_cTRDcEsBFpmNt8jfW1zJU4Nyss,46781
-llumo/exceptions.py,sha256=i3Qv4_g7XjRuho7-b7ybjw2bwSh_NhvICR6ZAgiLQX8,1944
-llumo/execution.py,sha256=x88wQV8eL99wNN5YtjFaAMCIfN1PdfQVlAZQb4vzgQ0,1413
-llumo/functionCalling.py,sha256=D5jYapu1rIvdIJNUYPYMTyhQ1H-6nkwoOLMi6eekfUE,7241
-llumo/helpingFuntions.py,sha256=RgWok8DoE1R-Tc0kJ9B5En6LEUEk5EvQU8iJiGPbUsw,21911
-llumo/models.py,sha256=YH-qAMnShmUpmKE2LQAzQdpRsaXkFSlOqMxHwU4zBUI,1560
-llumo/sockets.py,sha256=I2JO_eNEctRo_ikgvFVp5zDd-m0VDu04IEUhhsa1Tic,5950
-llumo-0.2.14b7.dist-info/licenses/LICENSE,sha256=tF9yAcfPV9xGT3ViWmC8hPvOo8BEk4ZICbUfcEo8Dlk,182
-llumo-0.2.14b7.dist-info/METADATA,sha256=kdeDmcNgV8uRyH7gXhhAqeb3se5U_Gqo3bA3Cf4SLlM,1521
-llumo-0.2.14b7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-llumo-0.2.14b7.dist-info/top_level.txt,sha256=d5zUTMI99llPtLRB8rtSrqELm_bOqX-bNC5IcwlDk88,6
-llumo-0.2.14b7.dist-info/RECORD,,