mistralai 1.9.11__py3-none-any.whl → 1.10.1__py3-none-any.whl

mistralai_gcp/fim.py

@@ -5,6 +5,7 @@ from mistralai_gcp import models, utils
 from mistralai_gcp._hooks import HookContext
 from mistralai_gcp.types import OptionalNullable, UNSET
 from mistralai_gcp.utils import eventstreaming
+from mistralai_gcp.utils.unmarshal_json_response import unmarshal_json_response
 from typing import Any, Mapping, Optional, Union
 
 
@@ -38,7 +39,7 @@ class Fim(BaseSDK):
 
         Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
@@ -105,6 +106,7 @@ class Fim(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_fim",
                 oauth2_scopes=[],
@@ -122,32 +124,23 @@ class Fim(BaseSDK):
                 http_res,
                 lambda raw: utils.unmarshal_json(raw, models.CompletionEvent),
                 sentinel="[DONE]",
+                client_ref=self,
             )
         if utils.match_response(http_res, "422", "application/json"):
             http_res_text = utils.stream_to_text(http_res)
-            response_data = utils.unmarshal_json(
-                http_res_text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res, http_res_text
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res, http_res_text)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "5XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        content_type = http_res.headers.get("Content-Type")
         http_res_text = utils.stream_to_text(http_res)
-        raise models.SDKError(
-            f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
-            http_res.status_code,
-            http_res_text,
-            http_res,
-        )
+        raise models.SDKError("Unexpected response received", http_res, http_res_text)
 
     async def stream_async(
         self,
@@ -176,7 +169,7 @@ class Fim(BaseSDK):
 
         Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
@@ -243,6 +236,7 @@ class Fim(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_fim",
                 oauth2_scopes=[],
@@ -260,32 +254,23 @@ class Fim(BaseSDK):
                 http_res,
                 lambda raw: utils.unmarshal_json(raw, models.CompletionEvent),
                 sentinel="[DONE]",
+                client_ref=self,
             )
         if utils.match_response(http_res, "422", "application/json"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            response_data = utils.unmarshal_json(
-                http_res_text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res, http_res_text
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res, http_res_text)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "5XX", "*"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        content_type = http_res.headers.get("Content-Type")
         http_res_text = await utils.stream_to_text_async(http_res)
-        raise models.SDKError(
-            f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
-            http_res.status_code,
-            http_res_text,
-            http_res,
-        )
+        raise models.SDKError("Unexpected response received", http_res, http_res_text)
 
     def complete(
         self,
@@ -314,7 +299,7 @@ class Fim(BaseSDK):
 
         FIM completion.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
@@ -381,6 +366,7 @@ class Fim(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="fim_completion_v1_fim_completions_post",
                 oauth2_scopes=[],
@@ -393,33 +379,22 @@ class Fim(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.FIMCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.FIMCompletionResponse], http_res
             )
         if utils.match_response(http_res, "422", "application/json"):
-            response_data = utils.unmarshal_json(
-                http_res.text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "5XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        content_type = http_res.headers.get("Content-Type")
-        http_res_text = utils.stream_to_text(http_res)
-        raise models.SDKError(
-            f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
-            http_res.status_code,
-            http_res_text,
-            http_res,
-        )
+        raise models.SDKError("Unexpected response received", http_res)
 
     async def complete_async(
         self,
@@ -448,7 +423,7 @@ class Fim(BaseSDK):
 
         FIM completion.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
@@ -515,6 +490,7 @@ class Fim(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="fim_completion_v1_fim_completions_post",
                 oauth2_scopes=[],
@@ -527,30 +503,19 @@ class Fim(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.FIMCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.FIMCompletionResponse], http_res
             )
         if utils.match_response(http_res, "422", "application/json"):
-            response_data = utils.unmarshal_json(
-                http_res.text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "5XX", "*"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        content_type = http_res.headers.get("Content-Type")
-        http_res_text = await utils.stream_to_text_async(http_res)
-        raise models.SDKError(
-            f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
-            http_res.status_code,
-            http_res_text,
-            http_res,
-        )
+        raise models.SDKError("Unexpected response received", http_res)

mistralai_gcp/httpclient.py

@@ -2,7 +2,6 @@
 
 # pyright: reportReturnType = false
 import asyncio
-from concurrent.futures import ThreadPoolExecutor
 from typing_extensions import Protocol, runtime_checkable
 import httpx
 from typing import Any, Optional, Union
@@ -116,21 +115,12 @@ def close_clients(
             pass
 
     if async_client is not None and not async_client_supplied:
-        is_async = False
         try:
-            asyncio.get_running_loop()
-            is_async = True
+            loop = asyncio.get_running_loop()
+            asyncio.run_coroutine_threadsafe(async_client.aclose(), loop)
         except RuntimeError:
-            pass
-
-        try:
-            # If this function is called in an async loop then start another
-            # loop in a separate thread to close the async http client.
-            if is_async:
-                with ThreadPoolExecutor(max_workers=1) as executor:
-                    future = executor.submit(asyncio.run, async_client.aclose())
-                    future.result()
-            else:
+            try:
                 asyncio.run(async_client.aclose())
-        except Exception:
-            pass
+            except RuntimeError:
+                # best effort
+                pass