xieyao/opencode-mobile
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix(cua): add Azure OpenAI support, fix max_tokens → max_completion_tokens, increase screencap timeout

Browse Source
This commit is contained in:
Ubuntu
2026-05-19 06:24:29 +00:00
parent e08a4a7a36
commit 7804b7a3b9
1 changed files with 12 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
scripts/android-cua-smoke.py
View File

@@ -52,7 +52,7 @@ import time
from pathlib import Path from pathlib import Path
try: try:
from openai import OpenAI from openai import OpenAI, AzureOpenAI
except ImportError: except ImportError:
sys.exit("openai package required: pip install openai") sys.exit("openai package required: pip install openai")
@@ -79,7 +79,7 @@ def screenshot_b64() -> str:
# Use exec-out for direct binary pipe (fastest, no intermediate file on device) # Use exec-out for direct binary pipe (fastest, no intermediate file on device)
result = subprocess.run( result = subprocess.run(
["adb", "exec-out", "screencap", "-p"], ["adb", "exec-out", "screencap", "-p"],
capture_output=True, timeout=10, capture_output=True, timeout=30,
) )
if result.returncode == 0 and len(result.stdout) > 100: if result.returncode == 0 and len(result.stdout) > 100:
return base64.b64encode(result.stdout).decode() return base64.b64encode(result.stdout).decode()
@@ -184,14 +184,14 @@ Rules:
""" """
def call_llm(client: OpenAI, model: str, system: str, history: list) -> str: def call_llm(client, model: str, system: str, history: list) -> str:
"""Call LLM via OpenAI-compatible API with retry on rate limit.""" """Call LLM via OpenAI-compatible API with retry on rate limit."""
for attempt in range(3): for attempt in range(3):
try: try:
response = client.chat.completions.create( response = client.chat.completions.create(
model=model, model=model,
messages=[{"role": "system", "content": system}] + history, messages=[{"role": "system", "content": system}] + history,
max_tokens=300, max_completion_tokens=300,
temperature=0, temperature=0,
) )
return response.choices[0].message.content.strip() return response.choices[0].message.content.strip()
@@ -205,7 +205,13 @@ def call_llm(client: OpenAI, model: str, system: str, history: list) -> str:
def make_client(model: str): def make_client(model: str):
"""Create OpenAI client. Supports OPENAI_API_KEY, GEMINI_API_KEY, XAI_API_KEY.""" """Create OpenAI client. Supports AZURE_OPENAI_*, OPENAI_API_KEY, GEMINI_API_KEY, XAI_API_KEY."""
if os.environ.get("AZURE_OPENAI_API_KEY"):
return AzureOpenAI(
api_key=os.environ["AZURE_OPENAI_API_KEY"],
azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
api_version=os.environ.get("AZURE_OPENAI_API_VERSION", "2024-08-01-preview"),
), model
if os.environ.get("OPENAI_API_KEY"): if os.environ.get("OPENAI_API_KEY"):
base = os.environ.get("OPENAI_BASE_URL") base = os.environ.get("OPENAI_BASE_URL")
return OpenAI(base_url=base) if base else OpenAI(), model return OpenAI(base_url=base) if base else OpenAI(), model
@@ -219,7 +225,7 @@ def make_client(model: str):
api_key=os.environ["GEMINI_API_KEY"], api_key=os.environ["GEMINI_API_KEY"],
base_url="https://generativelanguage.googleapis.com/v1beta/openai/", base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
), "gemini-2.0-flash" ), "gemini-2.0-flash"
sys.exit("Set OPENAI_API_KEY, XAI_API_KEY, or GEMINI_API_KEY") sys.exit("Set AZURE_OPENAI_API_KEY, OPENAI_API_KEY, XAI_API_KEY, or GEMINI_API_KEY")
def run_cua(goal: str, max_steps: int = 30, model: str = "gpt-4o", def run_cua(goal: str, max_steps: int = 30, model: str = "gpt-4o",