feat(cli): collapse regional duplicates; refresh Qwen catalog

Qwen and MiniMax each had two main-dropdown entries (intl + CN);
consolidate to one entry per provider and prompt for region as a
secondary step. Internal provider keys (qwen-cn, minimax-cn) and
endpoint routing unchanged. Add qwen3.6-flash to the Qwen catalog
and drop the version-less aliases (qwen-flash, qwen-plus) that
auto-shift their backing model per Alibaba's docs.

#758
This commit is contained in:
Yijia-Xiao
2026-05-11 04:16:11 +00:00
parent 0011b5ebf5
commit faaeebac70
3 changed files with 92 additions and 15 deletions

View File

@@ -559,6 +559,14 @@ def get_user_selections():
)
selected_llm_provider, backend_url = select_llm_provider()
# Providers with regional endpoints prompt for the region as a secondary
# step so the main dropdown stays clean (mainland China and international
# accounts cannot share API keys).
if selected_llm_provider == "qwen":
selected_llm_provider, backend_url = ask_qwen_region()
elif selected_llm_provider == "minimax":
selected_llm_provider, backend_url = ask_minimax_region()
# Step 7: Thinking agents
console.print(
create_question_box(

View File

@@ -237,10 +237,9 @@ def select_llm_provider() -> tuple[str, str | None]:
("Anthropic", "anthropic", "https://api.anthropic.com/"),
("xAI", "xai", "https://api.x.ai/v1"),
("DeepSeek", "deepseek", "https://api.deepseek.com"),
("Qwen", "qwen", "https://dashscope.aliyuncs.com/compatible-mode/v1"),
("Qwen", "qwen", "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"),
("GLM", "glm", "https://open.bigmodel.cn/api/paas/v4/"),
("MiniMax", "minimax", "https://api.minimax.io/v1"),
("MiniMax CN", "minimax-cn", "https://api.minimaxi.com/v1"),
("OpenRouter", "openrouter", "https://openrouter.ai/api/v1"),
("Azure OpenAI", "azure", None),
("Ollama", "ollama", "http://localhost:11434/v1"),
@@ -330,6 +329,60 @@ def ask_gemini_thinking_config() -> str | None:
).ask()
def ask_qwen_region() -> tuple[str, str]:
"""Ask which Qwen region (international vs China) to use.
Alibaba DashScope exposes two endpoints with separate accounts —
a key from one region does NOT authenticate against the other
(fixes #758). Returns (provider_key, backend_url).
"""
return questionary.select(
"Select Qwen region:",
choices=[
questionary.Choice(
"International — dashscope-intl.aliyuncs.com (uses DASHSCOPE_API_KEY)",
value=("qwen", "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"),
),
questionary.Choice(
"China — dashscope.aliyuncs.com (uses DASHSCOPE_CN_API_KEY)",
value=("qwen-cn", "https://dashscope.aliyuncs.com/compatible-mode/v1"),
),
],
style=questionary.Style([
("selected", "fg:cyan noinherit"),
("highlighted", "fg:cyan noinherit"),
("pointer", "fg:cyan noinherit"),
]),
).ask()
def ask_minimax_region() -> tuple[str, str]:
"""Ask which MiniMax region (global vs China) to use.
MiniMax exposes two endpoints with separate accounts — a key from
one region does NOT authenticate against the other. Returns
(provider_key, backend_url).
"""
return questionary.select(
"Select MiniMax region:",
choices=[
questionary.Choice(
"Global — api.minimax.io (uses MINIMAX_API_KEY)",
value=("minimax", "https://api.minimax.io/v1"),
),
questionary.Choice(
"China — api.minimaxi.com (uses MINIMAX_CN_API_KEY)",
value=("minimax-cn", "https://api.minimaxi.com/v1"),
),
],
style=questionary.Style([
("selected", "fg:cyan noinherit"),
("highlighted", "fg:cyan noinherit"),
("pointer", "fg:cyan noinherit"),
]),
).ask()
def ask_output_language() -> str:
"""Ask for report output language."""
choice = questionary.select(

View File

@@ -8,6 +8,31 @@ ModelOption = Tuple[str, str]
ProviderModeOptions = Dict[str, Dict[str, List[ModelOption]]]
# Shared model list for Qwen's global (dashscope-intl) and CN (dashscope) endpoints.
# Source: modelstudio.console.alibabacloud.com (Featured Models — Flagship + Cost-optimized).
#
# Only versioned IDs are exposed in the dropdown. The version-less aliases
# (qwen-plus, qwen-flash) are documented by Alibaba as auto-upgrading
# pointers ("backbone, latest, and snapshot ... have been upgraded to the
# Qwen3 series"), which means their behavior shifts when Alibaba rotates
# the backing model. Users who want a specific generation pick it
# explicitly; users who really want auto-latest can enter the alias via
# "Custom model ID".
_QWEN_MODELS: Dict[str, List[ModelOption]] = {
"quick": [
("Qwen 3.6 Flash - Latest fast, agentic coding + vision-language", "qwen3.6-flash"),
("Qwen 3.5 Flash - Previous-gen fast", "qwen3.5-flash"),
("Custom model ID", "custom"),
],
"deep": [
("Qwen 3.6 Plus - Flagship vision-language, agentic coding SOTA", "qwen3.6-plus"),
("Qwen 3.5 Plus - Previous-gen flagship", "qwen3.5-plus"),
("Qwen 3 Max - Specialized for agent programming + tool use", "qwen3-max"),
("Custom model ID", "custom"),
],
}
# Shared model list for MiniMax's global and CN endpoints (same IDs).
# Full official lineup per platform.minimax.io/docs/api-reference/text-openai-api.
# All M2.x models share a 204,800-token context window.
@@ -97,19 +122,10 @@ MODEL_OPTIONS: ProviderModeOptions = {
("Custom model ID", "custom"),
],
},
"qwen": {
"quick": [
("Qwen 3.5 Flash", "qwen3.5-flash"),
("Qwen Plus", "qwen-plus"),
("Custom model ID", "custom"),
],
"deep": [
("Qwen 3.6 Plus", "qwen3.6-plus"),
("Qwen 3.5 Plus", "qwen3.5-plus"),
("Qwen 3 Max", "qwen3-max"),
("Custom model ID", "custom"),
],
},
# Qwen: same model IDs across global (dashscope-intl) and China
# (dashscope) endpoints, so the two provider keys share one model list.
"qwen": _QWEN_MODELS,
"qwen-cn": _QWEN_MODELS,
"glm": {
"quick": [
("GLM-4.7", "glm-4.7"),