模型提供者¶
Provider 将模型后端适配到统一协议(LLMProvider:chat、
chat_stream、aclose)。内置的 Tongyi Provider 直接通过 httpx 与
DashScope 通信;OpenAI 与 Anthropic Provider 是可选扩展。
| Provider | 安装 | 说明 |
|---|---|---|
tongyi(默认) |
核心 | 通过 httpx 访问 DashScope 原生生成端点 |
openai |
pip install "dashscope-agenticx[provider-openai]" |
OpenAI 及兼容网关 |
anthropic |
pip install "dashscope-agenticx[provider-anthropic]" |
Anthropic Messages API |
自定义 Provider 可在 .agenticx/custom-extensions.toml 中声明,无需
编写 Python — 参见
工具、技能与专家。
路由遵循 Config.fallback_providers、MaaS 网关设置(workspace、
region、maas_api_key)以及由配置构建的各 Provider 档案。瞬时故障
由加固封装层重试;provider.switched 事件报告路由变化。
协议与共享类型¶
LLMProvider
¶
Bases: Protocol
aclose
async
¶
aclose() -> None
Source code in src/dashscope_agenticx/providers/base.py
57 58 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/base.py
60 61 62 63 64 65 66 | |
chat_stream
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/base.py
68 69 70 71 72 73 74 | |
ToolCall
dataclass
¶
ToolCall(
id: str,
name: str,
arguments: dict,
truncated: bool = False,
)
LLMResponse
dataclass
¶
LLMResponse(
content: str = "",
tool_calls: list[ToolCall] = list(),
reasoning_content: str = "",
usage: dict | None = None,
)
tool_calls
class-attribute
instance-attribute
¶
tool_calls: list[ToolCall] = field(default_factory=list)
LLMChunk
dataclass
¶
LLMChunk(
delta_content: str = "",
delta_reasoning_content: str = "",
tool_calls: list[ToolCall] = list(),
finish_reason: str | None = None,
usage: dict | None = None,
)
tool_calls
class-attribute
instance-attribute
¶
tool_calls: list[ToolCall] = field(default_factory=list)
内置 Provider¶
TongyiProvider
¶
TongyiProvider(
model: str = DEFAULT_MODEL,
api_key: str | None = None,
request_timeout: int = 60,
protocol: str = "openai",
base_url: str | None = None,
module: str = "app",
max_output_tokens: int | None = None,
thinking_budget: int | None = None,
)
Source code in src/dashscope_agenticx/providers/tongyi.py
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 | |
aclose
async
¶
aclose() -> None
Requests own their HTTP clients; no persistent transport is held.
Source code in src/dashscope_agenticx/providers/tongyi.py
328 329 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/tongyi.py
433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 | |
chat_stream
async
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/tongyi.py
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 | |
OpenAIProvider
¶
OpenAIProvider(
model: str = "gpt-4o",
api_key: str | None = None,
base_url: str | None = None,
module: str = "app",
)
Source code in src/dashscope_agenticx/providers/openai.py
18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 | |
client
instance-attribute
¶
client = AsyncOpenAI(
api_key=api_key,
base_url=base_url,
default_headers=dashscope_sdk_headers(
module, base_url=base_url or ""
),
)
aclose
async
¶
aclose() -> None
Source code in src/dashscope_agenticx/providers/openai.py
42 43 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/openai.py
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 | |
chat_stream
async
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/openai.py
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 | |
AnthropicProvider
¶
AnthropicProvider(
model: str = "claude-sonnet-4-20250514",
api_key: str | None = None,
base_url: str | None = None,
timeout: float = 120.0,
module: str = "app",
)
Source code in src/dashscope_agenticx/providers/anthropic.py
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 | |
client
instance-attribute
¶
client = anthropic.AsyncAnthropic(
api_key=api_key,
base_url=base_url,
timeout=timeout,
default_headers=dashscope_sdk_headers(
module, base_url=base_url or ""
),
)
aclose
async
¶
aclose() -> None
Source code in src/dashscope_agenticx/providers/anthropic.py
81 82 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/anthropic.py
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 | |
chat_stream
async
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/anthropic.py
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 | |
档案与路由¶
ProviderProfile
dataclass
¶
ProviderProfile(
name: str,
provider: str,
model: str,
api_key: str,
base_url: str | None = None,
timeout: float = 120.0,
protocol: str = "openai",
max_retries: int = 3,
module: str = "app",
)
Everything needed to talk to one model endpoint.
ProviderChain
¶
ProviderChain(profiles: list[ProviderProfile])
Bases: LLMProvider
Ordered list of ProviderProfiles; routes to the next on retryable failure.
Source code in src/dashscope_agenticx/providers/profile.py
309 310 311 | |
aclose
async
¶
aclose() -> None
Source code in src/dashscope_agenticx/providers/profile.py
313 314 315 316 317 318 319 320 321 322 323 324 325 326 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/profile.py
358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 | |
chat_stream
async
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/profile.py
385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 | |
build_profiles_from_config
¶
build_profiles_from_config(config) -> list[ProviderProfile]
Build primary and fallback profiles from AgenticX config.
Source code in src/dashscope_agenticx/providers/profile.py
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 | |
validate_key_for_profile
¶
validate_key_for_profile(
profile: ProviderProfile,
) -> list[str]
Return warnings if the API key looks mismatched for the endpoint.
Source code in src/dashscope_agenticx/providers/profile.py
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 | |
maas_gateway
¶
maas_gateway(config) -> tuple[str, str] | None
Return (base_url, api_key) for the new MaaS gateway.
None when workspace/region/maas_api_key are not all configured, in which case the caller falls back to the legacy gateway.
Source code in src/dashscope_agenticx/providers/profile.py
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 | |
warn_legacy_gateway
¶
warn_legacy_gateway(config) -> None
One-time deprecation notice for the legacy DashScope gateway.
Suppress with AGENTICX_DISABLE_GATEWAY_WARNING=1.
Source code in src/dashscope_agenticx/providers/profile.py
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 | |
加固(Hardening)¶
HardenedProvider
¶
HardenedProvider(
provider: LLMProvider,
max_retries: int = 3,
retry_delay: float = 4.0,
max_backoff: float = 16.0,
)
Wraps an LLMProvider with retries and empty-response recovery.
Source code in src/dashscope_agenticx/providers/hardening.py
202 203 204 205 206 207 208 209 210 211 212 | |
aclose
async
¶
aclose() -> None
Source code in src/dashscope_agenticx/providers/hardening.py
214 215 216 217 | |
chat
async
¶
chat(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> LLMResponse
Source code in src/dashscope_agenticx/providers/hardening.py
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 | |
chat_stream
async
¶
chat_stream(
messages: list[dict],
tools: list[dict] | None = None,
response_format: dict | None = None,
) -> AsyncIterator[LLMChunk]
Source code in src/dashscope_agenticx/providers/hardening.py
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 | |
RetriesExhausted
¶
RetriesExhausted(
error: BaseException, elapsed_sec: float = 0.0
)
Bases: RuntimeError
This layer already spent its whole retry budget on the wrapped error.
Keeps the original message so is_retryable_error still answers True
and ProviderChain keeps falling back to the next profile — the budget
that is exhausted is this layer's, not the chain's. What the type adds is
a signal for layers above the provider stack: re-attempting there pays
max_retries x request_timeout a second time, which against a stalled
endpoint measured 819s of silent wall clock from a 60s configured
timeout.
The type alone is not enough of a signal, though. A budget spent on a
stalled endpoint costs max_retries x request_timeout (268s measured);
the same budget spent on a refused connection costs only the ~28s of
backoff, and there an upper layer re-attempting is cheap and is what
rides out a transient blip. elapsed_sec lets the caller tell the two
apart instead of paying for the expensive case's protection in the cheap
case too.
Source code in src/dashscope_agenticx/providers/hardening.py
180 181 182 183 184 185 186 187 | |
is_retryable_error
¶
is_retryable_error(error: BaseException) -> bool
Classify an exception as retryable/fallback-worthy (bilingual).
Shared by HardenedProvider (retry) and ProviderChain (fallback) so both layers agree. Non-retryable patterns win over retryable ones.
Source code in src/dashscope_agenticx/providers/hardening.py
143 144 145 146 147 148 149 150 151 152 | |
account_error_hint
¶
account_error_hint(error: BaseException) -> str
Actionable hint for account/billing failures, else an empty string.
Source code in src/dashscope_agenticx/providers/hardening.py
120 121 122 123 124 125 126 127 128 129 | |
协议适配器¶
anthropic_to_openai_request
¶
anthropic_to_openai_request(
messages: list[dict],
system: str | None = None,
tools: list[dict] | None = None,
**kwargs
) -> dict
Convert Anthropic Messages API request to OpenAI Chat Completions format.
Anthropic format: { "system": "...", "messages": [{"role": "user", "content": "..."}], "tools": [{"name": "foo", "input_schema": {...}}] }
OpenAI format: { "messages": [ {"role": "system", "content": "..."}, {"role": "user", "content": "..."} ], "tools": [{"type": "function", "function": {"name": "foo", "parameters": {...}}}] }
Tool structures round-trip: assistant tool_use blocks (or the agent's
internal tool_calls key) become OpenAI tool_calls, and tool_result
blocks (or internal role: "tool" messages) become OpenAI tool
messages correlated by tool_call_id.
Source code in src/dashscope_agenticx/providers/adapter.py
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 | |
openai_to_anthropic_response
¶
openai_to_anthropic_response(openai_response: dict) -> dict
Convert OpenAI Chat Completions response to Anthropic Messages API format.
OpenAI format: { "choices": [{ "message": { "content": "...", "tool_calls": [{"id": "...", "function": {"name": "...", "arguments": "..."}}] }, "finish_reason": "stop" }], "usage": {"prompt_tokens": 10, "completion_tokens": 20, "total_tokens": 30} }
Anthropic format: { "id": "msg_...", "type": "message", "role": "assistant", "content": [ {"type": "text", "text": "..."}, {"type": "tool_use", "id": "...", "name": "...", "input": {...}} ], "stop_reason": "end_turn", "usage": {"input_tokens": 10, "output_tokens": 20} }
Source code in src/dashscope_agenticx/providers/adapter.py
216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 | |
anthropic_to_openai_response
¶
anthropic_to_openai_response(
anthropic_response: dict,
) -> dict
Convert an Anthropic Messages API response back to an OpenAI-style message dict: {"content": str, "tool_calls": [...]}.
Reverse of openai_to_anthropic_response's message part — tool_use blocks become tool_calls entries with JSON-string arguments.
Source code in src/dashscope_agenticx/providers/adapter.py
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 | |