Add claude-fable-5 to litellm proxy and bump Claude Code to 2.1.207#2437
Open
weishi-imbue wants to merge 4 commits into
Open
Add claude-fable-5 to litellm proxy and bump Claude Code to 2.1.207#2437weishi-imbue wants to merge 4 commits into
weishi-imbue wants to merge 4 commits into
Conversation
Fable pricing ($10/$50 per 1M, cache 1.25e-5/1e-6) mirrored from litellm's model_prices_and_context_window map, registered inline in both litellm configs and mngr_usage's MODEL_PRICING. Claude Code pin bumped at all five sites (Dockerfile ARG, snapshot script, CI workflows); the matching default-workspace-template pin bump lands in a separate PR. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
weishi-imbue
temporarily deployed
to
minds-ci-test
GitHub Actions
Inactive
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
weishi-imbue
temporarily deployed
to
minds-ci-test
GitHub Actions
Inactive
Problem: dev/changelog/mngr-fable.md described only the Claude Code CLI version bump, omitting the claude-fable-5 addition to the repo-root litellm_proxy/config.yaml, which is also a dev-project change. Fix: add a second bullet covering the config.yaml model addition and its inline pricing, matching how the prior model-lineup PR documented it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Problem: the mngr_robinhood Agent SDK resolves per-model pricing by family substring (haiku/sonnet/opus only), so claude-fable-5 sessions reported ResultMessage.total_cost_usd=None -- the same unpriced gap this branch closed in mngr_usage's token->USD table. Fix: add _FABLE_PRICING (input $10 / output $50 per 1M, cache write $12.50, cache read $1.00, matching litellm's price map) and a fable family entry, with a resolve test and a mngr_robinhood changelog entry. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
weishi-imbue
temporarily deployed
to
minds-ci-test
GitHub Actions
Inactive
Contributor
|
lgtm except for one required follow up change, see slack |
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
2 participants
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.
Makes Claude Fable 5 available in minds workspaces: the LiteLLM proxy learns to route and price
claude-fable-5, and the pinned Claude Code version moves to a Fable-capable release.Changes
LiteLLM proxy — add
claude-fable-5apps/modal_litellm/app.py(LITELLM_CONFIG) and its mirrorlitellm_proxy/config.yaml: new model entry with inline pricing mirrored from litellm'smodel_prices_and_context_windowmap — $10 / $50 per 1M input/output tokens, cache write 1.25e-5, cache read 1e-6 (2x Opus tier). Inline pricing keeps cost tracking correct regardless of the deployed litellm version's bundled price map;config_drift_test.pykeeps the two configs in sync.Usage pricing tables
libs/mngr_usage/.../pricing.py:anthropic/claude-fable-5entry so token-only usage sources price Fable instead of reporting it unpriced (mngr_usage_pricing_drift_test.pyenforces sync with modal_litellm).libs/mngr_robinhood/.../_agent_sdk/pricing.py: fable family entry so Agent SDK sessions reporttotal_cost_usdinstead ofNone, with a resolve-test assertion.Claude Code pin 2.1.160 → 2.1.207 (latest release; supports Fable) at all five sites: release Dockerfile
ARG,scripts/snapshot_minds_e2e_state.py,release-tests.yml(2),tmr-setupaction.Changelog entries for all five touched projects.
Verification
just test-offload— 15431 passed, 0 failed.dev-weishienv from this branch, minted a LiteLLM virtual key, and ranclaude -p --model claude-fable-5withANTHROPIC_BASE_URLpointed at the proxy (the exact workspace wire path). Request succeeded end-to-end and the spend log recordedanthropic/claude-fable-5, $0.735645, 60069 tokens— consistent with the cache-write rate (60k tokens x 1.25e-5 ≈ $0.75).Merge coordination
Companion PR imbue-ai/default-workspace-template#258 bumps
[agent_types.claude].versionto 2.1.207. The release testtest_claude_code_version_matches_default_workspace_template_pinrequires the two pins to match — merge both together (release tests don't run in CI, but agent provisioning fails on a version mismatch until both land).Rollout
Deployed proxies pick up the model list on their next
minds env deploy; staging/production deploys to be coordinated after merge. Workspaces on the SUBSCRIPTION/API_KEY AI providers only need the version bump.🤖 Generated with Claude Code