diff --git a/providers/volcengine-coding-plan/logo.svg b/providers/volcengine-coding-plan/logo.svg new file mode 100644 index 0000000000..5723862115 --- /dev/null +++ b/providers/volcengine-coding-plan/logo.svg @@ -0,0 +1,8 @@ + + Volcengine + + + + + + diff --git a/providers/volcengine-coding-plan/models/ark-code-latest.toml b/providers/volcengine-coding-plan/models/ark-code-latest.toml new file mode 100644 index 0000000000..03d6479384 --- /dev/null +++ b/providers/volcengine-coding-plan/models/ark-code-latest.toml @@ -0,0 +1,28 @@ +# Default Coding Plan alias that routes to the plan's latest coding model. +# Sources (accessed 2026-08-03): +# https://www.volcengine.com/docs/82379/1925114 +# https://www.volcengine.com/docs/82379/2556056 +name = "Ark Code Latest" +description = "Auto-routing alias for the latest coding model included in the Volcengine Ark Coding Plan" +family = "auto" +release_date = "2025-11" +last_updated = "2025-11" +attachment = false +reasoning = false +temperature = true +tool_call = true +open_weights = false + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 + +[limit] +context = 256_000 +output = 16_384 + +[modalities] +input = ["text"] +output = ["text"] diff --git a/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml b/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml new file mode 100644 index 0000000000..fee224a2e8 --- /dev/null +++ b/providers/volcengine-coding-plan/models/deepseek-v4-flash.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "deepseek/deepseek-v4-flash" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml b/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml new file mode 100644 index 0000000000..a5d2106730 --- /dev/null +++ b/providers/volcengine-coding-plan/models/deepseek-v4-pro.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "deepseek/deepseek-v4-pro" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml b/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml new file mode 100644 index 0000000000..bad9ed7309 --- /dev/null +++ b/providers/volcengine-coding-plan/models/doubao-seed-2.0-lite.toml @@ -0,0 +1,34 @@ +# ByteDance first-party coding model on its own Ark platform. Deep thinking is +# toggled with top-level `thinking.type` = enabled|disabled and returned as +# `reasoning_content`. Lite tier of the Doubao Seed 2.0 coding family. +# Sources (accessed 2026-08-03): +# https://www.volcengine.com/docs/82379/1925114 +# https://www.volcengine.com/docs/82379/1449737 +name = "Doubao Seed 2.0 Lite" +description = "ByteDance Doubao Seed 2.0 lite coding model for cost-efficient agentic software engineering" +family = "seed" +release_date = "2026-02" +last_updated = "2026-02" +attachment = false +reasoning = true +reasoning_options = [{ type = "toggle" }] +temperature = true +tool_call = true +open_weights = false + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 + +[limit] +context = 256_000 +output = 32_768 + +[modalities] +input = ["text"] +output = ["text"] diff --git a/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml b/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml new file mode 100644 index 0000000000..71dc38d021 --- /dev/null +++ b/providers/volcengine-coding-plan/models/doubao-seed-2.1-turbo.toml @@ -0,0 +1,34 @@ +# ByteDance first-party coding model on its own Ark platform. Deep thinking is +# toggled with top-level `thinking.type` = enabled|disabled and returned as +# `reasoning_content`. Turbo tier of the Doubao Seed 2.1 coding family. +# Sources (accessed 2026-08-03): +# https://www.volcengine.com/docs/82379/1925114 +# https://www.volcengine.com/docs/82379/1449737 +name = "Doubao Seed 2.1 Turbo" +description = "ByteDance Doubao Seed 2.1 turbo coding model for low-latency agentic software engineering" +family = "seed" +release_date = "2026-05" +last_updated = "2026-05" +attachment = false +reasoning = true +reasoning_options = [{ type = "toggle" }] +temperature = true +tool_call = true +open_weights = false + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 + +[limit] +context = 256_000 +output = 32_768 + +[modalities] +input = ["text"] +output = ["text"] diff --git a/providers/volcengine-coding-plan/models/glm-5.2.toml b/providers/volcengine-coding-plan/models/glm-5.2.toml new file mode 100644 index 0000000000..153cff5778 --- /dev/null +++ b/providers/volcengine-coding-plan/models/glm-5.2.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "zhipuai/glm-5.2" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/kimi-k2.6.toml b/providers/volcengine-coding-plan/models/kimi-k2.6.toml new file mode 100644 index 0000000000..69de65884a --- /dev/null +++ b/providers/volcengine-coding-plan/models/kimi-k2.6.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "moonshotai/kimi-k2.6" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml b/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml new file mode 100644 index 0000000000..8a816ff5ba --- /dev/null +++ b/providers/volcengine-coding-plan/models/kimi-k2.7-code.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "moonshotai/kimi-k2.7-code" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/minimax-m2.7.toml b/providers/volcengine-coding-plan/models/minimax-m2.7.toml new file mode 100644 index 0000000000..fea7b6837d --- /dev/null +++ b/providers/volcengine-coding-plan/models/minimax-m2.7.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "minimax/MiniMax-M2.7" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/models/minimax-m3.toml b/providers/volcengine-coding-plan/models/minimax-m3.toml new file mode 100644 index 0000000000..2ed5524621 --- /dev/null +++ b/providers/volcengine-coding-plan/models/minimax-m3.toml @@ -0,0 +1,15 @@ +# Toggle: thinking.type = enabled|disabled +# Coding Plan deep thinking is toggled with top-level `thinking.type` and +# returned as `reasoning_content`. +# https://www.volcengine.com/docs/82379/1925114 (accessed 2026-08-03) +base_model = "minimax/MiniMax-M3" +reasoning_options = [{ type = "toggle" }] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0 +output = 0 +cache_read = 0 +cache_write = 0 diff --git a/providers/volcengine-coding-plan/provider.toml b/providers/volcengine-coding-plan/provider.toml new file mode 100644 index 0000000000..6109c249c9 --- /dev/null +++ b/providers/volcengine-coding-plan/provider.toml @@ -0,0 +1,17 @@ +# OpenAI-compatible coding plan endpoint: POST +# https://ark.cn-beijing.volces.com/api/coding/v3/chat/completions (a Responses +# surface is also served). Deep thinking is toggled with top-level +# `thinking.type` = enabled|disabled and is returned as `reasoning_content` on +# the message or streamed delta. An Anthropic-compatible surface for +# Claude-style clients lives at /api/coding. Subscription usage must go through +# /api/coding/*; the pay-as-you-go /api/v3 endpoint bills separately. The plan +# bundles ByteDance Doubao Seed models plus DeepSeek, GLM, Kimi, and MiniMax. +# Sources: +# https://www.volcengine.com/docs/82379/1925114 +# https://www.volcengine.com/docs/82379/2556056 +# https://www.volcengine.com/activity/codingplan +name = "Volcengine Coding Plan (China)" +env = ["ARK_API_KEY"] +npm = "@ai-sdk/openai-compatible" +doc = "https://www.volcengine.com/docs/82379/1925114" +api = "https://ark.cn-beijing.volces.com/api/coding/v3"