feat(volcengine): add Volcano Engine (Doubao/Ark) provider support#895
Closed
dvrd wants to merge 18 commits intoRightNow-AI:mainfrom
Closed
feat(volcengine): add Volcano Engine (Doubao/Ark) provider support#895dvrd wants to merge 18 commits intoRightNow-AI:mainfrom
dvrd wants to merge 18 commits intoRightNow-AI:mainfrom
Conversation
Adds full support for the Alibaba Coding Plan (International) API endpoint (coding-intl.dashscope.aliyuncs.com/v1) as a new LLM provider. ## What's included - 8 models: qwen3.5-plus, qwen3-max-2026-01-23, qwen3-coder-plus, qwen3-coder-next, glm-4.7, glm-5, kimi-k2.5, minimax-m2.5 - Subscription pricing ($0.00 input / $0.00 output) for all 8 models - OpenAI-compatible driver wired via create_driver; prefix stripping handled upstream by strip_provider_prefix (underscore→hyphen normalization added for all providers) - resolve_catalog_model_id helper ensuring metering always resolves to the canonical prefixed catalog ID on all 3 call sites (streaming, non-streaming, compact_session) - detect_available_provider probe entry for auto-detection via ALIBABA_CODING_PLAN_API_KEY - Aliases: "alibaba-coding-plan" → qwen3.5-plus, "qwen3-coder" → qwen3-coder-plus - supports_vision: false for all 8 models (coding-specialized endpoint) - supports_function_calling: true for all 8 models - Provider accepts both "alibaba_coding_plan" and "alibaba-coding-plan" forms in config and create_driver - Full test coverage: driver creation, missing key, env isolation, catalog entries, alias resolution, metering $0 for all 8 models, strip_provider_prefix regression tests - docs/providers.md updated with model table, aliases, base URL, vision/function-calling flags Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(volcengine): 添加火山引擎(Doubao)和火山引擎编码计划支持
添加火山引擎(Doubao)和火山引擎编码计划作为新的AI提供商支持，包括：
- 在多个文件中添加VOLCENGINE_API_KEY环境变量配置
- 添加volcengine和volcengine_coding提供商信息
- 更新模型目录添加火山引擎相关模型
* feat(volcengine): 添加对火山引擎和豆包API的支持并更新模型目录
更新模型目录中的火山引擎编码计划模型，增加新模型并调整现有模型的参数
* feat(tui): 添加并调整火山引擎提供商配置
将火山引擎相关配置移至列表更靠前的位置，并确保其配置完整
* refactor(volcengine): standardize display name for Volcano Engine provider
Updated the display name for the Volcano Engine provider across multiple files to remove the "(Doubao)" suffix for consistency.
* fix(volcengine): resolve duplicate model ID and env var guard issues
- Rename doubao-seed-code under volcengine_coding provider to
doubao-seed-code-ark to eliminate ambiguous model ID lookup
- Add comment marking third-party Ark marketplace models in catalog
- Guard empty env vars in launcher.rs and welcome.rs detect_provider()
using .ok().filter(|v| !v.is_empty()).is_some() instead of .is_ok()
- Add volcengine and volcengine_coding to unknown-provider error message
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): resolve duplicate model IDs, clarify separators, add provider tests
- Prefix ark-marketplace models with "ark/" to avoid ID collision with
native minimax, zhipu, and moonshot provider entries (FIX 1)
- Add comment explaining dot vs hyphen separator difference between
volcengine_coding (coding/v3) and volcengine (v3) endpoints (FIX 2)
- Include "doubao" alias in unknown-provider error message (FIX 3)
- Add "volcengine" probe to detect_available_provider() (FIX 4)
- Add test_provider_defaults_volcengine, _doubao_alias, and
_volcengine_coding tests (FIX 5)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): display names, ordering, and documentation comments
- Add display field to ProviderInfo in wizard.rs; all providers now show
human-readable names (e.g. "Volcano Engine" / "Volcano Engine (Coding)")
instead of raw identifiers
- Swap volcengine/volcengine_coding order to match init_wizard.rs and catalog
- Document "doubao" dual-alias intent in runtime model_catalog.rs
- Document cn-beijing regional hardcoding with base_url override note
- Document zero pricing for Ark-routed third-party models
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): correct provider probe order, display names, and alias comment
- Swap volcengine/volcengine_coding probe order in detect_available_provider so
the general volcengine provider is auto-selected when VOLCENGINE_API_KEY is set
- Align volcengine_coding display name to "Volcano Engine (Coding Plan)" in both
init_wizard.rs and wizard.rs
- Add sync-warning comment on the doubao alias in model_catalog.rs
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): remove volcengine_coding from auto-detect probe list and add wizard hints
volcengine_coding shares the same VOLCENGINE_API_KEY as volcengine, making it
permanently unreachable in auto-detection. Remove it from PROBE_ORDER so only
volcengine is auto-detected; volcengine_coding remains selectable via the wizard.
Also add descriptive hints for both volcengine entries in the init wizard
("ByteDance Ark platform" / "ByteDance Ark — coding models").
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): strip ark/ prefix from Ark marketplace model IDs and fix provider order
Move ark/ prefix to aliases so catalog lookups still resolve but the
bare model name is forwarded to the Ark API (which rejects the prefix).
Also reorder volcengine before volcengine_coding in provider_list().
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): strip ark/ prefix before sending model id to Ark API
strip_provider_prefix now removes the ark/ prefix so that model IDs
like ark/minimax-m2.5 are sent as minimax-m2.5 to the Ark API.
Also extends test_ark_alias_resolution to cover find_model paths.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(runtime): gate ark/ prefix stripping to volcengine providers only
Previously strip_provider_prefix stripped the ark/ prefix for all
providers, which could corrupt model IDs from other providers. Now
the strip only fires when provider is "volcengine" or "volcengine_coding".
Also adds two unit tests verifying the guarded and unguarded cases.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): add doubao provider alias to ark/ strip guard and extend tests
- Add "doubao" to the ark/ prefix strip condition in agent_loop.rs so
users with provider = "doubao" get correct model IDs sent to the API
- Add two new tests: volcengine and doubao ark/ stripping
- Update volcengine wizard hints with region info (cn-beijing)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(model_catalog): resolve Ark marketplace model ID collisions
The four Ark marketplace models under volcengine_coding had bare IDs
(minimax-m2.5, glm-4.7, deepseek-v3.2, kimi-k2.5) that collided with
native provider entries. Since find_model does a linear scan returning
the first match, ark/ lookups were resolving to the wrong provider.
Fix: make ark/ the canonical ID for each colliding entry, keeping the
bare name as an alias only for deepseek-v3.2 (no collision exists there).
Update test_ark_alias_resolution to assert on the new canonical ark/ IDs
and verify native provider entries remain unaffected.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(volcengine): clarify ark/ catalog-vs-wire comment and rename test
---------
Co-authored-by: gaoyiman <gaoyiman@bytedance.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
添加火山引擎(Doubao)和火山引擎编码计划作为新的AI提供商支持，包括： - 在多个文件中添加VOLCENGINE_API_KEY环境变量配置 - 添加volcengine和volcengine_coding提供商信息 - 更新模型目录添加火山引擎相关模型
更新模型目录中的火山引擎编码计划模型，增加新模型并调整现有模型的参数
将火山引擎相关配置移至列表更靠前的位置，并确保其配置完整
…vider Updated the display name for the Volcano Engine provider across multiple files to remove the "(Doubao)" suffix for consistency.
… provider tests - Prefix ark-marketplace models with "ark/" to avoid ID collision with native minimax, zhipu, and moonshot provider entries (FIX 1) - Add comment explaining dot vs hyphen separator difference between volcengine_coding (coding/v3) and volcengine (v3) endpoints (FIX 2) - Include "doubao" alias in unknown-provider error message (FIX 3) - Add "volcengine" probe to detect_available_provider() (FIX 4) - Add test_provider_defaults_volcengine, _doubao_alias, and _volcengine_coding tests (FIX 5) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add display field to ProviderInfo in wizard.rs; all providers now show human-readable names (e.g. "Volcano Engine" / "Volcano Engine (Coding)") instead of raw identifiers - Swap volcengine/volcengine_coding order to match init_wizard.rs and catalog - Document "doubao" dual-alias intent in runtime model_catalog.rs - Document cn-beijing regional hardcoding with base_url override note - Document zero pricing for Ark-routed third-party models Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…as comment - Swap volcengine/volcengine_coding probe order in detect_available_provider so the general volcengine provider is auto-selected when VOLCENGINE_API_KEY is set - Align volcengine_coding display name to "Volcano Engine (Coding Plan)" in both init_wizard.rs and wizard.rs - Add sync-warning comment on the doubao alias in model_catalog.rs Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… and add wizard hints
volcengine_coding shares the same VOLCENGINE_API_KEY as volcengine, making it
permanently unreachable in auto-detection. Remove it from PROBE_ORDER so only
volcengine is auto-detected; volcengine_coding remains selectable via the wizard.
Also add descriptive hints for both volcengine entries in the init wizard
("ByteDance Ark platform" / "ByteDance Ark — coding models").
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… fix provider order Move ark/ prefix to aliases so catalog lookups still resolve but the bare model name is forwarded to the Ark API (which rejects the prefix). Also reorder volcengine before volcengine_coding in provider_list(). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
strip_provider_prefix now removes the ark/ prefix so that model IDs like ark/minimax-m2.5 are sent as minimax-m2.5 to the Ark API. Also extends test_ark_alias_resolution to cover find_model paths. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…tend tests - Add "doubao" to the ark/ prefix strip condition in agent_loop.rs so users with provider = "doubao" get correct model IDs sent to the API - Add two new tests: volcengine and doubao ark/ stripping - Update volcengine wizard hints with region info (cn-beijing) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The four Ark marketplace models under volcengine_coding had bare IDs (minimax-m2.5, glm-4.7, deepseek-v3.2, kimi-k2.5) that collided with native provider entries. Since find_model does a linear scan returning the first match, ark/ lookups were resolving to the wrong provider. Fix: make ark/ the canonical ID for each colliding entry, keeping the bare name as an alias only for deepseek-v3.2 (no collision exists there). Update test_ark_alias_resolution to assert on the new canonical ark/ IDs and verify native provider entries remain unaffected. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…nd add constants
- FIX 1: detect_auth now filters empty strings via .ok().filter(|v| !v.is_empty()).is_some()
for both primary key and gemini/codex fallbacks, preventing blank env vars from
triggering AuthStatus::Configured
- FIX 2: volcengine_coding skips /models probe in test_api_key (coding endpoint does not
expose a standard OpenAI-compatible models list); falls through to return true
- FIX 4: add test_doubao_dual_role_consistency asserting VOLCENGINE_BASE_URL contains
volces.com and that catalog.find_model("doubao").provider == "volcengine"
- FIX 5: add VOLCENGINE_PROVIDER_ID and VOLCENGINE_CODING_PROVIDER_ID constants to
openfang-types for use as single source of truth
- FIX 6: disambiguate display names — "Doubao Seed 2.0 Code (Coding Plan)" and
"Doubao Seed Code (Standard Plan)" to avoid UI confusion between the two plans
- FIX 7: update volcengine_coding wizard hint to mention shared VOLCENGINE_API_KEY
and Ark Coding endpoint
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Provider count increased from 38 to 39 after rebasing onto main which added alibaba_coding_plan. Update the hardcoded assertion to match.
- Add "doubao" to known_providers() and update count to 40 - Add create_driver smoke tests for "doubao" alias (direct key + no-key error path) - Adopt VOLCENGINE_PROVIDER_ID / VOLCENGINE_CODING_PROVIDER_ID constants in strip_provider_prefix guard (eliminates dead-constant warning) - Remove stray double blank line in model_catalog.rs
…r error message - Add DOUBAO_PROVIDER_ID constant to openfang-types and use it in strip_provider_prefix guard (all three volcengine aliases now use named constants, eliminating bare string literals) - Expand unknown-provider error message to list all known providers (qwen-code, kimi_coding, zhipu_coding, zai, zai_coding, qianfan, moonshot, lemonade were previously omitted)
Author
|
Closing — changes will be merged into our fork first. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
volcengineandvolcengine_codingproviders backed by ByteDance's Ark API (ark.cn-beijing.volces.com)doubaoas a provider alias routing to the volcengine Ark endpointChanges
Core
openfang-types:VOLCENGINE_BASE_URL,VOLCENGINE_CODING_BASE_URL,VOLCENGINE_PROVIDER_ID,VOLCENGINE_CODING_PROVIDER_ID,DOUBAO_PROVIDER_IDconstantsopenfang-runtime/drivers/mod.rs: provider defaults, known providers (40 total), auto-detect probe,doubaoalias armopenfang-runtime/model_catalog.rs: 12 models across two providers;ark/namespace for Ark marketplace models to avoid ID collisions with native providers (minimax, zhipu, moonshot, deepseek)openfang-runtime/agent_loop.rs:strip_provider_prefixstripsark/catalog namespace before sending to Ark API; uses named constants for all three volcengine provider IDsTUI / CLI
init_wizard.rs+wizard.rs: both providers added with hints and correct env varlauncher.rs+welcome.rs:VOLCENGINE_API_KEYdetectionmain.rs: provider list, key validation;volcengine_codingprobe bypassed (shares key withvolcengine, coding plan requires explicit selection)Bug fixes caught during review
ark/canonical namespace!v.is_empty()) applied consistently indetect_authvolcengine_codingremoved from auto-detect probeark/stripping gated to volcengine providers only (not unconditional)Tests
test_provider_defaults_volcengine/test_provider_defaults_volcengine_codingtest_ark_alias_resolution— verifiesark/minimax-m2.5etc. resolve correctlytest_doubao_dual_role_consistency— verifies doubao alias coherencetest_doubao_alias_driver_with_direct_key/test_doubao_alias_driver_no_key_errorstest_strip_ark_catalog_prefix_for_volcengine_coding+ negative case for non-volcengine providersCloses #873