Repository navigation
Add built-in ClinePass and Ollama Cloud provider presets - #12
Merged
codeking-ai merged 3 commits intoOct 15, 2026
Merged
codeking-ai merged 3 commits into
codeking-ai merged 3 commits into
Conversation
…x to presets - Load user-defined provider presets from ~/.cligate/provider-presets.json at module init; validated entries merge additively after built-ins and become first-class (add-key form, model mapping, discovery, protocols, pricing). Invalid entries are skipped with warnings; built-in ids are protected from collision. - Migrate Zhipu/Moonshot/MiniMax from bespoke classes to declarative presets served by the generic OpenAI-compatible provider. Existing api-keys.json entries keep working (same ids, base URLs, pricing). New preset field validateModel: Zhipu key validation probes a 1-token chat request instead of GET /models (which Zhipu does not serve). - Zhipu/Moonshot/MiniMax gain tier maps + static model lists: non-native models now remap via tiers instead of passing through verbatim. - Fix settings-page normalizeModelMappingProvider: stop folding zhipu/moonshot/minimax into 'openai' — provider identity in the mapping UI is now 1:1 with runtime resolveModel(providerType, ...). - Add docs/CUSTOM_PROVIDERS.md and unit tests for the loader, the migration, and the normalization fix. Co-Authored-By: Claude Code <noreply@anthropic.com>
…2026) Model ids verified against OpenRouter vendor listings (z-ai, moonshotai, minimax) and official pricing pages (platform.kimi.ai, platform.minimax.io, docs.z.ai). - zhipu: glm-5.3 family (flagship glm-5.3, standard glm-5.2, fast glm-5.3-flash); drop retired glm-5-turbo/glm-4-plus/glm-4-air* - moonshot: kimi-k3 flagship, kimi-k2.6 standard, kimi-k2.7-code highspeed fast; drop retired kimi-k2.5/kimi-k2-thinking - minimax: MiniMax-M3 flagship, M2.7 standard; fix M2.7-highspeed input price (0.60, was 0.30) - pricing tables synced: new cacheRead rates and 2026 prices; glm-4.7-flash/glm-4-flash are free Co-Authored-By: Claude Code <noreply@anthropic.com>
- clinepass: Cline's flat subscription for open-weight models via api.cline.bot/api/v1 (OpenAI chat-compatible; ids carry the mandatory cline-pass/ slug prefix, so a single native prefix covers passthrough). Pricing entries are the upstream reference rates used for quota measurement, not per-token billing. - ollama-cloud: hosted models at ollama.com/v1 (the local Ollama runtime keeps the `ollama` id — this preset is the cloud-only path). Passthrough combines exact family prefixes (gpt-oss, glm, kimi, deepseek, ...) with a `:` match for tag-style ids like gpt-oss:120b, so explicitly requested catalog ids are never tier-remapped. - i18n labels (EN/ZH), docs built-in id list, preset test coverage. Co-Authored-By: Claude Code <noreply@anthropic.com>
Owner
|
Thanks! Verified the ClinePass catalog and reference pricing against https://docs.cline.bot/getting-started/clinepass — model slugs and rates match exactly. Preset tests pass. Merging (after #11). One issue to fix in a follow-up: several Ollama Cloud model ids don't match the live catalog.
If Ollama doesn't alias the untagged names, any request remapped to the reasoning tier will fail with model-not-found. Please update Minor: since cloud model availability shifts often (ClinePass already deprecated several models), it may be worth relying on |
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
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
Follow-up to #11 (user-defined provider presets): adds two popular subscriptions as built-in presets, so they work out of the box without a hand-written JSON file.
clinepass— ClinePass (Cline subscription)Cline's flat subscription ($9.99/mo) for open-weight models (GLM, Kimi, DeepSeek, MiMo, MiniMax, Qwen) served from
https://api.cline.bot/api/v1. OpenAI chat-compatible only upstream (no native Anthropic endpoint) — Claude Code still works through the generic provider's Anthropic⇄chat bridge. Every model id carries the mandatorycline-pass/slug prefix, so passthrough is a single prefix rule.pricingholds the upstream reference rates (used for quota measurement — the subscription is flat).ollama-cloud— Ollama CloudHosted models at
https://ollama.com/v1(OpenAI-compatible,/modelsavailable). The local Ollama runtime keeps the existingollamaid — this preset is the cloud-only path, henceollama-cloud. Cloud model ids mix bare vendor names (glm-5.3,kimi-k3) and tag-style ones (gpt-oss:120b), so passthrough combines exact family prefixes (gpt-oss— deliberately notgpt, or incoming gpt-5.x from Codex would stop remapping — plus glm/kimi/deepseek/minimax/gemma/mistral/nemotron/qwen/llama/phi/granite) with a:substring match.Both presets land automatically in the add-key form, model mapping (own tier maps), discovery, and pricing via the existing data-driven preset pipeline — no new provider classes.
Test plan
tests/unit/provider-presets.test.js: extended with clinepass/ollama-cloud cases — native passthrough (cline-pass/*, bare +:-tagged ids), tier remap of non-native ids (claude-sonnet-4-6→ tier model;gpt-5.4NOT swallowed by a baregptprefix), both addable with the Anthropic bridge and no native/responses.tests/unit/provider-presets.test.jsformat list +tests/unit/provider-user-presets.test.jsbuilt-in id list updated.npm run test:unit: no new failures vs master.🤖 Generated with Claude Code