oimwiodev/ChatMock
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add GPT-5-Codex support (#37)

Browse Source
This commit is contained in:
Maciej Wilczyński
2025-09-16 14:58:41 +02:00
committed by GitHub
parent 2f23cd5a89
commit 77d60fe321
10 changed files with 189 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
chatmock/app.py
View File

@@ -2,7 +2,7 @@ from __future__ import annotations
from flask import Flask, jsonify
from .config import BASE_INSTRUCTIONS
from .config import BASE_INSTRUCTIONS, GPT5_CODEX_INSTRUCTIONS
from .http import build_cors_headers
from .routes_openai import openai_bp
from .routes_ollama import ollama_bp
@@ -26,6 +26,7 @@ def create_app(
REASONING_COMPAT=reasoning_compat,
DEBUG_MODEL=debug_model,
BASE_INSTRUCTIONS=BASE_INSTRUCTIONS,
GPT5_CODEX_INSTRUCTIONS=GPT5_CODEX_INSTRUCTIONS,
EXPOSE_REASONING_MODELS=bool(expose_reasoning_models),
DEFAULT_WEB_SEARCH=bool(default_web_search),
)

35
chatmock/config.py
View File

@@ -10,26 +10,37 @@ CLIENT_ID_DEFAULT = os.getenv("CHATGPT_LOCAL_CLIENT_ID") or "app_EMoamEEZ73f0CkX
CHATGPT_RESPONSES_URL = "https://chatgpt.com/backend-api/codex/responses"
def read_base_instructions() -> str:
def _read_prompt_text(filename: str) -> str | None:
candidates = [
Path(__file__).parent.parent / "prompt.md",
Path(__file__).parent / "prompt.md",
Path(getattr(sys, "_MEIPASS", "")) / "prompt.md" if getattr(sys, "_MEIPASS", None) else None,
Path.cwd() / "prompt.md",
Path(__file__).parent.parent / filename,
Path(__file__).parent / filename,
Path(getattr(sys, "_MEIPASS", "")) / filename if getattr(sys, "_MEIPASS", None) else None,
Path.cwd() / filename,
]
for p in candidates:
if not p:
for candidate in candidates:
if not candidate:
continue
try:
if p.exists():
content = p.read_text(encoding="utf-8")
if candidate.exists():
content = candidate.read_text(encoding="utf-8")
if isinstance(content, str) and content.strip():
return content
except Exception:
continue
raise FileNotFoundError(
"Failed to read prompt.md; expected adjacent to package or CWD."
)
return None
def read_base_instructions() -> str:
content = _read_prompt_text("prompt.md")
if content is None:
raise FileNotFoundError("Failed to read prompt.md; expected adjacent to package or CWD.")
return content
def read_gpt5_codex_instructions(fallback: str) -> str:
content = _read_prompt_text("prompt_gpt5_codex.md")
return content if isinstance(content, str) and content.strip() else fallback
BASE_INSTRUCTIONS = read_base_instructions()
GPT5_CODEX_INSTRUCTIONS = read_gpt5_codex_instructions(BASE_INSTRUCTIONS)

32
chatmock/routes_ollama.py
View File

@@ -7,7 +7,7 @@ from typing import Any, Dict, List
from flask import Blueprint, Response, current_app, jsonify, make_response, request, stream_with_context
from .config import BASE_INSTRUCTIONS
from .config import BASE_INSTRUCTIONS, GPT5_CODEX_INSTRUCTIONS
from .http import build_cors_headers
from .reasoning import build_reasoning_param, extract_reasoning_from_model_name
from .transform import convert_ollama_messages, normalize_ollama_tools
@@ -18,6 +18,15 @@ from .utils import convert_chat_messages_to_responses_input, convert_tools_chat_
ollama_bp = Blueprint("ollama", __name__)
def _instructions_for_model(model: str) -> str:
base = current_app.config.get("BASE_INSTRUCTIONS", BASE_INSTRUCTIONS)
if model == "gpt-5-codex":
codex = current_app.config.get("GPT5_CODEX_INSTRUCTIONS") or GPT5_CODEX_INSTRUCTIONS
if isinstance(codex, str) and codex.strip():
return codex
return base
_OLLAMA_FAKE_EVAL = {
"total_duration": 8497226791,
"load_duration": 1747193958,
@@ -33,19 +42,19 @@ def ollama_tags() -> Response:
if bool(current_app.config.get("VERBOSE")):
print("IN GET /api/tags")
expose_variants = bool(current_app.config.get("EXPOSE_REASONING_MODELS"))
model_ids = [
"gpt-5",
*(
model_ids = ["gpt-5", "gpt-5-codex"]
if expose_variants:
model_ids.extend(
[
"gpt-5-high",
"gpt-5-medium",
"gpt-5-low",
"gpt-5-minimal",
"gpt-5-codex-high",
"gpt-5-codex-medium",
"gpt-5-codex-low",
]
if expose_variants
else []
),
]
)
models = []
for model_id in model_ids:
models.append(
@@ -184,10 +193,11 @@ def ollama_chat() -> Response:
input_items = convert_chat_messages_to_responses_input(messages)
model_reasoning = extract_reasoning_from_model_name(model)
normalized_model = normalize_model_name(model)
upstream, error_resp = start_upstream_request(
normalize_model_name(model),
normalized_model,
input_items,
instructions=BASE_INSTRUCTIONS,
instructions=_instructions_for_model(normalized_model),
tools=tools_responses,
tool_choice=tool_choice,
parallel_tool_calls=parallel_tool_calls,
@@ -231,7 +241,7 @@ def ollama_chat() -> Response:
)
created_at = datetime.datetime.utcnow().strftime("%Y-%m-%dT%H:%M:%SZ")
model_out = model if isinstance(model, str) and model.strip() else normalize_model_name(model)
model_out = model if isinstance(model, str) and model.strip() else normalized_model
if stream_req:
def _gen():

37
chatmock/routes_openai.py
View File

@@ -6,7 +6,7 @@ from typing import Any, Dict, List
from flask import Blueprint, Response, current_app, jsonify, make_response, request
from .config import BASE_INSTRUCTIONS
from .config import BASE_INSTRUCTIONS, GPT5_CODEX_INSTRUCTIONS
from .http import build_cors_headers
from .reasoning import apply_reasoning_to_message, build_reasoning_param, extract_reasoning_from_model_name
from .upstream import normalize_model_name, start_upstream_request
@@ -21,6 +21,15 @@ from .utils import (
openai_bp = Blueprint("openai", __name__)
def _instructions_for_model(model: str) -> str:
base = current_app.config.get("BASE_INSTRUCTIONS", BASE_INSTRUCTIONS)
if model == "gpt-5-codex":
codex = current_app.config.get("GPT5_CODEX_INSTRUCTIONS") or GPT5_CODEX_INSTRUCTIONS
if isinstance(codex, str) and codex.strip():
return codex
return base
@openai_bp.route("/v1/chat/completions", methods=["POST"])
def chat_completions() -> Response:
verbose = bool(current_app.config.get("VERBOSE"))
@@ -125,7 +134,7 @@ def chat_completions() -> Response:
upstream, error_resp = start_upstream_request(
model,
input_items,
instructions=BASE_INSTRUCTIONS,
instructions=_instructions_for_model(model),
tools=tools_responses,
tool_choice=tool_choice,
parallel_tool_calls=parallel_tool_calls,
@@ -327,7 +336,7 @@ def completions() -> Response:
upstream, error_resp = start_upstream_request(
model,
input_items,
instructions=BASE_INSTRUCTIONS,
instructions=_instructions_for_model(model),
reasoning_param=reasoning_param,
)
if error_resp is not None:
@@ -424,18 +433,16 @@ def completions() -> Response:
@openai_bp.route("/v1/models", methods=["GET"])
def list_models() -> Response:
expose_variants = bool(current_app.config.get("EXPOSE_REASONING_MODELS"))
data = []
if expose_variants:
variant_ids = [
"gpt-5",
"gpt-5-high",
"gpt-5-medium",
"gpt-5-low",
"gpt-5-minimal",
]
data = [{"id": mid, "object": "model", "owned_by": "owner"} for mid in variant_ids]
else:
data = [{"id": "gpt-5", "object": "model", "owned_by": "owner"}]
model_groups = [
("gpt-5", ["high", "medium", "low", "minimal"]),
("gpt-5-codex", ["high", "medium", "low"]),
]
model_ids: List[str] = []
for base, efforts in model_groups:
model_ids.append(base)
if expose_variants:
model_ids.extend([f"{base}-{effort}" for effort in efforts])
data = [{"id": mid, "object": "model", "owned_by": "owner"} for mid in model_ids]
models = {"object": "list", "data": data}
resp = make_response(jsonify(models), 200)
for k, v in build_cors_headers().items():

3
chatmock/upstream.py
View File

@@ -31,6 +31,9 @@ def normalize_model_name(name: str | None, debug_model: str | None = None) -> st
"gpt5": "gpt-5",
"gpt-5-latest": "gpt-5",
"gpt-5": "gpt-5",
"gpt5-codex": "gpt-5-codex",
"gpt-5-codex": "gpt-5-codex",
"gpt-5-codex-latest": "gpt-5-codex",
"codex": "codex-mini-latest",
"codex-mini": "codex-mini-latest",
"codex-mini-latest": "codex-mini-latest",