v2.17.0 (#822)
Co-authored-by: Clément Sirieix <clement.sirieix@mistral.ai> Co-authored-by: Guillaume LE GOFF <guillaume.lgf@gmail.com> Co-authored-by: Hdandria <henri.dandria@mistral.ai> Co-authored-by: Ivana Dunisijevic <ivana.dunisijevic@mistral.ai> Co-authored-by: Jean Burellier <sheplu@users.noreply.github.com> Co-authored-by: Mathias Gesbert <mathias.gesbert@mistral.ai> Co-authored-by: Mert Unsal <mert.unsal@mistral.ai> Co-authored-by: Michel Thomazo <51709227+michelTho@users.noreply.github.com> Co-authored-by: Paul VEZIA <166131032+le-codeur-rapide@users.noreply.github.com> Co-authored-by: Pierre Rossinès <pierre.rossines@mistral.ai> Co-authored-by: Val <102326092+vdeva@users.noreply.github.com> Co-authored-by: Vincent G <10739306+VinceOPS@users.noreply.github.com> Co-authored-by: renovate-mistral[bot] <253709520+renovate-mistral[bot]@users.noreply.github.com> Co-authored-by: Mistral Vibe <vibe@mistral.ai>
This commit is contained in:
parent
564a14365e
commit
6bedf271ce
223 changed files with 10533 additions and 6947 deletions
|
|
@ -8,6 +8,8 @@ import threading
|
|||
import time
|
||||
from typing import TypedDict, cast
|
||||
|
||||
from tests.constants import CHAT_COMPLETIONS_PATH
|
||||
|
||||
|
||||
class StreamOptionsPayload(TypedDict, total=False):
|
||||
include_usage: bool
|
||||
|
|
@ -85,6 +87,104 @@ class StreamingMockServer:
|
|||
),
|
||||
]
|
||||
|
||||
@staticmethod
|
||||
def _merge_tool_call_delta(
|
||||
tool_calls_by_index: dict[int, dict[str, object]],
|
||||
delta_tool_call: object,
|
||||
*,
|
||||
fallback_index: int | None,
|
||||
) -> int | None:
|
||||
if not isinstance(delta_tool_call, dict):
|
||||
return fallback_index
|
||||
|
||||
index = delta_tool_call.get("index")
|
||||
if not isinstance(index, int):
|
||||
index = fallback_index
|
||||
if index is None:
|
||||
index = len(tool_calls_by_index)
|
||||
|
||||
tool_call = tool_calls_by_index.setdefault(index, {"index": index})
|
||||
for key in ("id", "type"):
|
||||
if value := delta_tool_call.get(key):
|
||||
tool_call[key] = value
|
||||
|
||||
function_delta = delta_tool_call.get("function")
|
||||
if not isinstance(function_delta, dict):
|
||||
return index
|
||||
|
||||
function = tool_call.setdefault("function", {})
|
||||
if not isinstance(function, dict):
|
||||
function = {}
|
||||
tool_call["function"] = function
|
||||
|
||||
if name := function_delta.get("name"):
|
||||
function["name"] = name
|
||||
if arguments := function_delta.get("arguments"):
|
||||
function["arguments"] = f"{function.get('arguments', '')}{arguments}"
|
||||
return index
|
||||
|
||||
@staticmethod
|
||||
def _completion_response_from_chunks(
|
||||
chunks: list[StreamChunk],
|
||||
) -> dict[str, object]:
|
||||
content_parts: list[str] = []
|
||||
tool_calls_by_index: dict[int, dict[str, object]] = {}
|
||||
active_tool_call_index: int | None = None
|
||||
finish_reason: object = "stop"
|
||||
usage: object = {"prompt_tokens": 3, "completion_tokens": 4}
|
||||
created = 123
|
||||
|
||||
for chunk in chunks:
|
||||
chunk_created = chunk.get("created")
|
||||
if isinstance(chunk_created, int):
|
||||
created = chunk_created
|
||||
if chunk_usage := chunk.get("usage"):
|
||||
usage = chunk_usage
|
||||
choices = chunk.get("choices")
|
||||
if not isinstance(choices, list) or not choices:
|
||||
continue
|
||||
choice = choices[0]
|
||||
if not isinstance(choice, dict):
|
||||
continue
|
||||
if choice.get("finish_reason") is not None:
|
||||
finish_reason = choice.get("finish_reason")
|
||||
delta = choice.get("delta")
|
||||
if not isinstance(delta, dict):
|
||||
continue
|
||||
if content := delta.get("content"):
|
||||
content_parts.append(str(content))
|
||||
if delta_tool_calls := delta.get("tool_calls"):
|
||||
if isinstance(delta_tool_calls, list):
|
||||
for delta_tool_call in delta_tool_calls:
|
||||
active_tool_call_index = (
|
||||
StreamingMockServer._merge_tool_call_delta(
|
||||
tool_calls_by_index,
|
||||
delta_tool_call,
|
||||
fallback_index=active_tool_call_index,
|
||||
)
|
||||
)
|
||||
|
||||
message: dict[str, object] = {
|
||||
"role": "assistant",
|
||||
"content": "".join(content_parts),
|
||||
}
|
||||
tool_calls = [
|
||||
tool_calls_by_index[index] for index in sorted(tool_calls_by_index)
|
||||
]
|
||||
if tool_calls:
|
||||
message["tool_calls"] = tool_calls
|
||||
|
||||
return {
|
||||
"id": "mock-id",
|
||||
"object": "chat.completion",
|
||||
"created": created,
|
||||
"model": "mock-model",
|
||||
"choices": [
|
||||
{"index": 0, "message": message, "finish_reason": finish_reason}
|
||||
],
|
||||
"usage": usage,
|
||||
}
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
|
|
@ -110,7 +210,7 @@ class StreamingMockServer:
|
|||
return
|
||||
|
||||
def do_POST(self) -> None:
|
||||
if self.path != "/v1/chat/completions":
|
||||
if self.path != CHAT_COMPLETIONS_PATH:
|
||||
self.send_response(404)
|
||||
self.end_headers()
|
||||
return
|
||||
|
|
@ -125,17 +225,28 @@ class StreamingMockServer:
|
|||
parent.requests.append(payload)
|
||||
request_index = len(parent.requests) - 1
|
||||
|
||||
self.send_response(200)
|
||||
self.send_header("Content-Type", "text/event-stream")
|
||||
self.send_header("Cache-Control", "no-cache")
|
||||
self.end_headers()
|
||||
|
||||
chunks = (
|
||||
parent._chunk_factory(request_index, payload)
|
||||
if parent._chunk_factory is not None
|
||||
else parent._stream_chunks()
|
||||
)
|
||||
|
||||
if not payload.get("stream"):
|
||||
response = parent._completion_response_from_chunks(chunks)
|
||||
response_body = json.dumps(response, ensure_ascii=False).encode()
|
||||
self.send_response(200)
|
||||
self.send_header("Content-Type", "application/json")
|
||||
self.send_header("Content-Length", str(len(response_body)))
|
||||
self.end_headers()
|
||||
self.wfile.write(response_body)
|
||||
self.wfile.flush()
|
||||
return
|
||||
|
||||
self.send_response(200)
|
||||
self.send_header("Content-Type", "text/event-stream")
|
||||
self.send_header("Cache-Control", "no-cache")
|
||||
self.end_headers()
|
||||
|
||||
for chunk in chunks:
|
||||
data = json.dumps(chunk, ensure_ascii=False)
|
||||
self.wfile.write(f"data: {data}\n\n".encode())
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue