pygeai 0.6.0b7__py3-none-any.whl → 0.6.0b10__py3-none-any.whl

pygeai/chat/clients.py

@@ -1,12 +1,14 @@
 import json
 from json import JSONDecodeError
+from pathlib import Path
 from typing import List, Dict, Optional, Union, Generator
 
 from pygeai import logger
-from pygeai.chat.endpoints import CHAT_V1, GENERATE_IMAGE_V1
+from pygeai.chat.endpoints import CHAT_V1, GENERATE_IMAGE_V1, RESPONSES_V1
 from pygeai.core.base.clients import BaseClient
 from pygeai.core.common.exceptions import InvalidAPIResponseException
 from pygeai.core.utils.validators import validate_status_code
+from pygeai.core.utils.parsers import parse_json_response
 
 
 class ChatClient(BaseClient):
@@ -15,8 +17,8 @@ class ChatClient(BaseClient):
         response = self.api_service.post(
             endpoint=CHAT_V1
         )
-        result = response.json()
-        return result
+        validate_status_code(response)
+        return parse_json_response(response, "chat")
 
     def chat_completion(
             self,
@@ -40,7 +42,8 @@ class ChatClient(BaseClient):
             stream_options: Optional[Dict] = None,
             store: Optional[bool] = None,
             metadata: Optional[Dict] = None,
-            user: Optional[str] = None
+            user: Optional[str] = None,
+            reasoning_effort: Optional[str] = None
     ) -> Union[dict, str, Generator[str, None, None]]:
         """
         Generates a chat completion response using the specified model and parameters.
@@ -50,8 +53,8 @@ class ChatClient(BaseClient):
         :param messages: List[Dict[str, str]] - A list of messages to include in the chat completion. Each message should be a dictionary with
             the following structure:
                 {
-                    "role": "string",  # Possible values: "user", "system", "assistant", or others supported by the model
-                    "content": "string"  # The content of the message
+                    "role": "string",
+                    "content": "string"
                 } (Required)
         :param stream: bool - Whether the response should be streamed. Possible values:
             - False: Returns the complete response as a dictionary or string (default).
@@ -78,6 +81,9 @@ class ChatClient(BaseClient):
         :param store: Optional[bool] - Whether to store the output for model distillation or evals. (Optional)
         :param metadata: Optional[Dict] - Up to 16 key-value pairs to attach to the object. (Optional)
         :param user: Optional[str] - A unique identifier for the end-user to monitor abuse. (Optional)
+        :param reasoning_effort: Optional[str] - Controls the depth of reasoning applied by supported models.
+            Possible values: "low", "medium", "high". Supported by OpenAI models from version 5,
+            Claude models from version 4.1, and Gemini models from version 2.0. (Optional)
         :return: Union[dict, str, Generator[str, None, None]] - For non-streaming (stream=False), returns a dictionary containing the chat completion
             result or a string if JSON decoding fails. For streaming (stream=True), returns a generator yielding content strings extracted from the
             streaming response chunks.
@@ -87,19 +93,16 @@ class ChatClient(BaseClient):
             'messages': messages,
             'stream': stream
         }
-        if temperature:
+        if temperature is not None:
             data['temperature'] = temperature
 
         if max_tokens:
             data['max_completion_tokens'] = max_tokens
 
-        if thread_id:
-            data['threadId'] = thread_id
-
-        if frequency_penalty:
+        if frequency_penalty is not None:
             data['frequency_penalty'] = frequency_penalty
 
-        if presence_penalty:
+        if presence_penalty is not None:
             data['presence_penalty'] = presence_penalty
 
         if variables is not None and any(variables):
@@ -141,6 +144,9 @@ class ChatClient(BaseClient):
         if user is not None:
             data['user'] = user
 
+        if reasoning_effort is not None:
+            data['reasoning_effort'] = reasoning_effort
+
         headers = {}
         if thread_id:
             headers["saia-conversation-id"] = thread_id
@@ -160,12 +166,13 @@ class ChatClient(BaseClient):
                 data=data,
                 headers=headers
             )
-            try:
-                result = response.json()
-                logger.debug(f"Chat completion result: {result}")
-                return result
-            except JSONDecodeError as e:
-                raise InvalidAPIResponseException(f"Unable to process chat request: {response.text}")
+            validate_status_code(response)
+
+            result = parse_json_response(response, "process chat request")
+
+            logger.debug(f"Chat completion result: {result}")
+
+            return result
 
     def stream_chat_generator(self, response) -> Generator[str, None, None]:
         """
@@ -195,6 +202,34 @@ class ChatClient(BaseClient):
         except Exception as e:
             raise InvalidAPIResponseException(f"Unable to process streaming chat response: {e}")
 
+    def stream_response_generator(self, response) -> Generator[str, None, None]:
+        """
+        Processes a streaming response from the Responses API and yields content strings.
+
+        :param response: The streaming response from the API.
+        :return: Generator[str, None, None] - Yields content strings extracted from streaming chunks.
+        """
+        try:
+            for line in response:
+                if line.startswith("data:"):
+                    chunk = line[5:].strip()
+                    if chunk == "[DONE]":
+                        break
+                    try:
+                        json_data = json.loads(chunk)
+                        if (
+                                json_data.get("choices")
+                                and len(json_data["choices"]) > 0
+                                and "delta" in json_data["choices"][0]
+                                and "content" in json_data["choices"][0]["delta"]
+                        ):
+                            content = json_data["choices"][0]["delta"]["content"]
+                            yield content
+                    except JSONDecodeError as e:
+                        continue
+        except Exception as e:
+            raise InvalidAPIResponseException(f"Unable to process streaming response: {e}")
+
     def generate_image(
             self,
             model: str,
@@ -212,7 +247,7 @@ class ChatClient(BaseClient):
         :param n: int - Number of images to generate (1-10, depending on the model). (Required)
         :param quality: str - Rendering quality, e.g., "high". (Required)
         :param size: str - Image dimensions, e.g., "1024x1024". (Required)
-        :param aspect_ratio: Optional[str] - Relationship between image’s width and height, e.g., "1:1", "9:16", "16:9", "3:4", "4:3". (Optional)
+        :param aspect_ratio: Optional[str] - Relationship between image's width and height, e.g., "1:1", "9:16", "16:9", "3:4", "4:3". (Optional)
         :return: dict - The API response containing the generated image data.
         :raises InvalidAPIResponseException: If the API response cannot be processed.
         """
@@ -234,9 +269,140 @@ class ChatClient(BaseClient):
             data=data
         )
 
-        try:
-            result = response.json()
-            logger.debug(f"Image generation result: {result}")
-            return result
-        except JSONDecodeError as e:
-            raise InvalidAPIResponseException(f"Unable to process image generation request: {response.text}")
+        validate_status_code(response)
+
+        result = parse_json_response(response, "generate image")
+        logger.debug(f"Image generation result: {result}")
+        return result
+
+    def get_response(
+            self,
+            model: str,
+            input: str,
+            files: Optional[List[str]] = None,
+            tools: Optional[List[Dict]] = None,
+            tool_choice: Optional[Union[str, Dict]] = None,
+            temperature: Optional[float] = None,
+            max_output_tokens: Optional[int] = None,
+            top_p: Optional[float] = None,
+            metadata: Optional[Dict] = None,
+            user: Optional[str] = None,
+            instructions: Optional[str] = None,
+            reasoning: Optional[Dict] = None,
+            truncation: Optional[str] = None,
+            parallel_tool_calls: Optional[bool] = None,
+            store: Optional[bool] = None,
+            stream: bool = False
+    ) -> Union[dict, str, Generator[str, None, None]]:
+        """
+        Generates a response using the Responses API with support for images and PDF files.
+
+        :param model: str - The model specification, e.g., "openai/o1-pro". (Required)
+        :param input: str - The user input text. (Required)
+        :param files: Optional[List[str]] - List of file paths (images or PDFs) to include in the request. (Optional)
+        :param tools: Optional[List[Dict]] - A list of tools (e.g., functions) the model may call. (Optional)
+        :param tool_choice: Optional[Union[str, Dict]] - Controls which tool is called (e.g., "none", "auto", or specific tool). (Optional)
+        :param temperature: Optional[float] - Controls the randomness of the response. (Optional)
+        :param max_output_tokens: Optional[int] - The maximum number of tokens to generate in the response. (Optional)
+        :param top_p: Optional[float] - Nucleus sampling parameter. (Optional)
+        :param metadata: Optional[Dict] - Up to 16 key-value pairs to attach to the object. (Optional)
+        :param user: Optional[str] - A unique identifier for the end-user. (Optional)
+        :param instructions: Optional[str] - Additional instructions for the model. (Optional)
+        :param reasoning: Optional[Dict] - Reasoning configuration, e.g., {"effort": "medium"}. (Optional)
+        :param truncation: Optional[str] - Truncation strategy, e.g., "disabled". (Optional)
+        :param parallel_tool_calls: Optional[bool] - Whether to enable parallel tool calls. (Optional)
+        :param store: Optional[bool] - Whether to store the output. (Optional)
+        :param stream: bool - Whether the response should be streamed. Possible values:
+            - False: Returns the complete response as a dictionary or string (default).
+            - True: Returns a generator yielding content strings as they are received. (Optional)
+        :return: Union[dict, str, Generator[str, None, None]] - For non-streaming (stream=False), returns a dictionary containing the response
+            result or a string if JSON decoding fails. For streaming (stream=True), returns a generator yielding content strings extracted from the
+            streaming response chunks.
+        :raises InvalidAPIResponseException: If the API response cannot be processed.
+        """
+        data = {
+            'model': model,
+            'input': input,
+            'stream': stream
+        }
+
+        if temperature is not None:
+            data['temperature'] = temperature
+
+        if max_output_tokens is not None:
+            data['max_output_tokens'] = max_output_tokens
+
+        if top_p is not None:
+            data['top_p'] = top_p
+
+        if tools is not None:
+            data['tools'] = tools
+
+        if tool_choice is not None:
+            data['tool_choice'] = tool_choice
+
+        if metadata is not None:
+            data['metadata'] = metadata
+
+        if user is not None:
+            data['user'] = user
+
+        if instructions is not None:
+            data['instructions'] = instructions
+
+        if reasoning is not None:
+            data['reasoning'] = reasoning
+
+        if truncation is not None:
+            data['truncation'] = truncation
+
+        if parallel_tool_calls is not None:
+            data['parallel_tool_calls'] = parallel_tool_calls
+
+        if store is not None:
+            data['store'] = store
+
+        logger.debug(f"Generating response with data: {data}")
+
+        if files:
+            if stream:
+                raise InvalidAPIResponseException("Streaming is not supported when uploading files")
+            
+            file_handles = []
+            try:
+                files_dict = {}
+                for idx, file_path in enumerate(files):
+                    path = Path(file_path)
+                    if not path.is_file():
+                        raise FileNotFoundError(f"File not found: {file_path}")
+                    
+                    file_handle = path.open("rb")
+                    file_handles.append(file_handle)
+                    files_dict[f"file{idx}"] = file_handle
+
+                response = self.api_service.post_files_multipart(
+                    endpoint=RESPONSES_V1,
+                    data=data,
+                    files=files_dict
+                )
+            finally:
+                for fh in file_handles:
+                    fh.close()
+        else:
+            if stream:
+                response = self.api_service.stream_post(
+                    endpoint=RESPONSES_V1,
+                    data=data
+                )
+                return self.stream_response_generator(response)
+            else:
+                response = self.api_service.post(
+                    endpoint=RESPONSES_V1,
+                    data=data
+                )
+
+        validate_status_code(response)
+        result = parse_json_response(response, "get response")
+        logger.debug(f"Response result: {result}")
+
+        return result

pygeai/chat/endpoints.py

@@ -1,3 +1,4 @@
 CHAT_V1 = "/chat"                           # POST Chat endpoint
 CHAT_COMPLETION_V1 = "/chat/completion"     # POST Chat completion endpoint
-GENERATE_IMAGE_V1 = "/images"               # POST Generate image
+GENERATE_IMAGE_V1 = "/images"               # POST Generate image
+RESPONSES_V1 = "/responses"                 # POST Responses API endpoint

pygeai/cli/commands/chat.py

@@ -52,6 +52,7 @@ def get_chat_completion(option_list: list):
     store = None
     metadata = None
     user = None
+    reasoning_effort = None
 
     for option_flag, option_arg in option_list:
         if option_flag.name == "model":
@@ -124,6 +125,8 @@ def get_chat_completion(option_list: list):
                 raise WrongArgumentError("metadata must be a valid JSON object")
         if option_flag.name == "user":
             user = option_arg
+        if option_flag.name == "reasoning_effort":
+            reasoning_effort = option_arg
 
     messages = get_messages(message_list)
 
@@ -151,7 +154,8 @@ def get_chat_completion(option_list: list):
         stream_options=stream_options,
         store=store,
         metadata=metadata,
-        user=user
+        user=user,
+        reasoning_effort=reasoning_effort
     )
     if stream:
         Console.write_stdout("Streaming chat completion:")
@@ -295,6 +299,14 @@ chat_completion_options = [
         "Optional string identifier for the end-user to monitor abuse.",
         True
     ),
+    Option(
+        "reasoning_effort",
+        ["--reasoning-effort"],
+        "Optional string to control the depth of reasoning applied by supported models. "
+        "Possible values: 'low', 'medium', 'high'. Supported by OpenAI models from version 5, "
+        "Claude models from version 4.1, and Gemini models from version 2.0.",
+        True
+    ),
 ]
 
 
@@ -649,6 +661,211 @@ generate_image_options = [
 ]
 
 
+def get_response(option_list: list):
+    model = None
+    input_text = None
+    files = None
+    tools = None
+    tool_choice = None
+    temperature = None
+    max_output_tokens = None
+    top_p = None
+    metadata = None
+    user = None
+    instructions = None
+    reasoning = None
+    truncation = None
+    parallel_tool_calls = None
+    store = None
+    stream = False
+
+    for option_flag, option_arg in option_list:
+        if option_flag.name == "model":
+            model = option_arg
+        if option_flag.name == "input":
+            input_text = option_arg
+        if option_flag.name == "files":
+            try:
+                files = json.loads(option_arg) if option_arg else None
+                if files and not isinstance(files, list):
+                    raise WrongArgumentError("files must be a JSON array of file paths")
+            except json.JSONDecodeError:
+                raise WrongArgumentError("files must be a valid JSON array")
+        if option_flag.name == "tools":
+            try:
+                tools = json.loads(option_arg) if option_arg else None
+            except json.JSONDecodeError:
+                raise WrongArgumentError("tools must be a valid JSON array")
+        if option_flag.name == "tool_choice":
+            try:
+                tool_choice = json.loads(option_arg) if option_arg else None
+            except json.JSONDecodeError:
+                tool_choice = option_arg
+        if option_flag.name == "temperature":
+            temperature = float(option_arg) if option_arg is not None else None
+        if option_flag.name == "max_output_tokens":
+            max_output_tokens = int(option_arg) if option_arg is not None else None
+        if option_flag.name == "top_p":
+            top_p = float(option_arg) if option_arg else None
+        if option_flag.name == "metadata":
+            try:
+                metadata = json.loads(option_arg) if option_arg else None
+            except json.JSONDecodeError:
+                raise WrongArgumentError("metadata must be a valid JSON object")
+        if option_flag.name == "user":
+            user = option_arg
+        if option_flag.name == "instructions":
+            instructions = option_arg
+        if option_flag.name == "reasoning":
+            try:
+                reasoning = json.loads(option_arg) if option_arg else None
+            except json.JSONDecodeError:
+                raise WrongArgumentError("reasoning must be a valid JSON object")
+        if option_flag.name == "truncation":
+            truncation = option_arg
+        if option_flag.name == "parallel_tool_calls":
+            parallel_tool_calls = get_boolean_value(option_arg) if option_arg else None
+        if option_flag.name == "store":
+            store = get_boolean_value(option_arg) if option_arg else None
+        if option_flag.name == "stream":
+            if option_arg:
+                stream = get_boolean_value(option_arg)
+
+    if not (model and input_text):
+        raise MissingRequirementException("Cannot get response without specifying model and input")
+
+    client = ChatClient()
+    try:
+        result = client.get_response(
+            model=model,
+            input=input_text,
+            files=files,
+            tools=tools,
+            tool_choice=tool_choice,
+            temperature=temperature,
+            max_output_tokens=max_output_tokens,
+            top_p=top_p,
+            metadata=metadata,
+            user=user,
+            instructions=instructions,
+            reasoning=reasoning,
+            truncation=truncation,
+            parallel_tool_calls=parallel_tool_calls,
+            store=store,
+            stream=stream
+        )
+        if stream:
+            Console.write_stdout("Streaming response:")
+            for chunk in result:
+                Console.write_stdout(f"{chunk}", end="")
+                sys.stdout.flush()
+            Console.write_stdout()
+        else:
+            Console.write_stdout(f"Response result: \n{json.dumps(result, indent=2)}\n")
+    except Exception as e:
+        logger.error(f"Error getting response: {e}")
+        Console.write_stderr(f"Failed to get response: {e}")
+
+
+response_options = [
+    Option(
+        "model",
+        ["--model", "-m"],
+        "The model specification, e.g., 'openai/o1-pro'.",
+        True
+    ),
+    Option(
+        "input",
+        ["--input", "-i"],
+        "The user input text.",
+        True
+    ),
+    Option(
+        "files",
+        ["--files", "-f"],
+        "JSON array of file paths (images or PDFs) to include in the request, e.g., '[\"image.jpg\", \"doc.pdf\"]'.",
+        True
+    ),
+    Option(
+        "tools",
+        ["--tools"],
+        "Optional JSON array of tools (e.g., functions) the model may call.",
+        True
+    ),
+    Option(
+        "tool_choice",
+        ["--tool-choice"],
+        "Optional string (e.g., \"none\", \"auto\") or JSON object to control which tool is called.",
+        True
+    ),
+    Option(
+        "temperature",
+        ["--temperature", "--temp"],
+        "Float value to set randomness of the response (between 0 and 2).",
+        True
+    ),
+    Option(
+        "max_output_tokens",
+        ["--max-output-tokens"],
+        "Integer value to set max tokens in the output.",
+        True
+    ),
+    Option(
+        "top_p",
+        ["--top-p"],
+        "Optional float value for nucleus sampling (between 0 and 1).",
+        True
+    ),
+    Option(
+        "metadata",
+        ["--metadata"],
+        "Optional JSON object with up to 16 key-value pairs to attach to the object.",
+        True
+    ),
+    Option(
+        "user",
+        ["--user"],
+        "Optional string identifier for the end-user.",
+        True
+    ),
+    Option(
+        "instructions",
+        ["--instructions"],
+        "Optional additional instructions for the model.",
+        True
+    ),
+    Option(
+        "reasoning",
+        ["--reasoning"],
+        "Optional JSON object for reasoning configuration, e.g., {\"effort\": \"medium\"}.",
+        True
+    ),
+    Option(
+        "truncation",
+        ["--truncation"],
+        "Optional truncation strategy, e.g., \"disabled\".",
+        True
+    ),
+    Option(
+        "parallel_tool_calls",
+        ["--parallel-tool-calls"],
+        "Optional boolean to enable parallel tool calls. Possible values: 0: OFF; 1: ON",
+        True
+    ),
+    Option(
+        "store",
+        ["--store"],
+        "Optional boolean to store the output. Possible values: 0: OFF; 1: ON",
+        True
+    ),
+    Option(
+        "stream",
+        ["--stream"],
+        "Whether to stream the response. Possible values: 0: OFF; 1: ON",
+        True
+    ),
+]
+
 chat_commands = [
     Command(
         "help",
@@ -695,5 +912,14 @@ chat_commands = [
         [],
         generate_image_options
     ),
+    Command(
+        "response",
+        ["response", "resp"],
+        "Get a response using the Responses API with support for images and PDFs",
+        get_response,
+        ArgumentsEnum.REQUIRED,
+        [],
+        response_options
+    ),
 
 ]

pygeai/cli/commands/embeddings.py

@@ -1,5 +1,7 @@
+import json
 from pygeai.cli.commands import Command, Option, ArgumentsEnum
 from pygeai.cli.commands.builders import build_help_text
+from pygeai.cli.commands.common import get_boolean_value
 from pygeai.cli.texts.help import EMBEDDINGS_HELP_TEXT
 from pygeai.core.common.exceptions import MissingRequirementException, WrongArgumentError
 from pygeai.core.embeddings.clients import EmbeddingsClient
@@ -22,6 +24,7 @@ def generate_embeddings(option_list: list):
     input_type = None
     timeout = None
     cache = None
+    preview = True
     input_list = list()
 
     for option_flag, option_arg in option_list:
@@ -32,18 +35,23 @@ def generate_embeddings(option_list: list):
         if option_flag.name == "encoding_format":
             encoding_format = option_arg
         if option_flag.name == "dimensions":
-            dimensions = option_arg
+            try:
+                dimensions = int(option_arg)
+            except (ValueError, TypeError):
+                raise WrongArgumentError("dimensions must be an integer")
         if option_flag.name == "user":
             user = option_arg
         if option_flag.name == "input_type":
             input_type = option_arg
         if option_flag.name == "timeout":
-            timeout = option_arg
+            try:
+                timeout = int(option_arg)
+            except (ValueError, TypeError):
+                raise WrongArgumentError("timeout must be an integer")
         if option_flag.name == "cache":
-            if not str(option_arg).isdigit() or int(option_arg) not in [0, 1]:
-                raise WrongArgumentError("If specified, cache must be 0 or 1")
-
-            cache = bool(int(option_arg))
+            cache = get_boolean_value(option_arg)
+        if option_flag.name == "preview":
+            preview = get_boolean_value(option_arg)
 
     if not (model and any(input_list)):
         raise MissingRequirementException("Cannot generate embeddings without specifying model and at least one input")
@@ -59,7 +67,40 @@ def generate_embeddings(option_list: list):
         timeout=timeout,
         cache=cache
     )
-    Console.write_stdout(f"Embeddings detail: \n{result}")
+    
+    output = {
+        "model": result.get("model"),
+        "object": result.get("object"),
+        "embeddings_count": len(result.get("data", [])),
+        "usage": result.get("usage"),
+        "data": []
+    }
+    
+    for item in result.get("data", []):
+        embedding_data = item.get("embedding")
+        if isinstance(embedding_data, list):
+            embedding_info = {
+                "index": item.get("index"),
+                "dimensions": len(embedding_data),
+                "object": item.get("object")
+            }
+            if preview:
+                embedding_info["preview"] = embedding_data[:5] if len(embedding_data) > 5 else embedding_data
+            else:
+                embedding_info["embedding"] = embedding_data
+        else:
+            embedding_info = {
+                "index": item.get("index"),
+                "object": item.get("object"),
+                "format": "base64"
+            }
+            if preview:
+                embedding_info["preview"] = str(embedding_data)[:50] + "..." if len(str(embedding_data)) > 50 else embedding_data
+            else:
+                embedding_info["embedding"] = embedding_data
+        output["data"].append(embedding_info)
+    
+    Console.write_stdout(json.dumps(output, indent=2))
 
 
 generate_embeddings_options = [
@@ -109,10 +150,17 @@ generate_embeddings_options = [
     Option(
         "cache",
         ["--cache"],
-        "Enable X-Saia-Cache-Enabled to cache the embeddings for the model; it applies by Organization/Project."
+        "Enable X-Saia-Cache-Enabled to cache the embeddings for the model; it applies by Organization/Project. "
         "1 to set to True and 0 to false. 0 is default",
         True
     ),
+    Option(
+        "preview",
+        ["--preview"],
+        "Control embedding display in output. 1 (default) shows a preview (first 5 values for float, 50 chars for base64). "
+        "0 shows the full embedding vector. Use 0 to get complete embeddings for further processing",
+        True
+    ),
 
 ]