feat(litellm): actual per-call cost + cache accounting for the open-weight backend - #66

Merged

CarlesUIPath merged 18 commits into

mainfrom

feat/openweight-actual-cost

Jul 31, 2026

Commits on Jul 30, 2026

feat(litellm): actual per-call cost + cache accounting for the open-weight backend

CarlesUIPath
and
claude
committed
fix(litellm): pin each open-weight model to a vetted provider set (no silent fallback)

CarlesUIPath
and
claude
committed
test(litellm): cover config shape, join ordering, and defensive cost branches

CarlesUIPath
and
claude
committed
Merge branch 'main' into feat/openweight-actual-cost
CarlesUIPath
authored
fix(litellm): gate cost_log_tags on agent capability, not route (fixes non-Claude crash)

CarlesUIPath
and
claude
committed
fix(litellm): proxy-authoritative token buckets + all-priced gate + transactional join

CarlesUIPath
and
claude
committed
fix(litellm): per-attempt cost-log scoping + single run-id accessor + no-match warning

CarlesUIPath
and
claude
committed
fix(litellm): sanitize cost headers, reject non-finite cost, drop debug scaffolding

CarlesUIPath
and
claude
committed
docs(litellm): document LITELLM_COST_LOG wiring + correct the reconciliation-cost contract

CarlesUIPath
and
claude
committed
refactor(litellm): stream the cost log + de-duplicate the OpenRouter config comment

CarlesUIPath
and
claude
committed
refactor(litellm): cut per-message distribution; turn-level join + per-call audit record

CarlesUIPath
and
claude
committed
docs(litellm): correct the cost contract after cutting per-message distribution

CarlesUIPath
and
claude
committed
feat(evalboard): per-call cost/cache table from provider_call_costs (replaces inline)

CarlesUIPath
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(litellm): actual per-call cost + cache accounting for the open-weight backend - #66

feat(litellm): actual per-call cost + cache accounting for the open-weight backend#66
CarlesUIPath merged 18 commits into
mainfrom
feat/openweight-actual-cost

Commits on Jul 30, 2026

feat(litellm): actual per-call cost + cache accounting for the open-weight backend

fix(litellm): pin each open-weight model to a vetted provider set (no silent fallback)

test(litellm): cover config shape, join ordering, and defensive cost branches

Merge branch 'main' into feat/openweight-actual-cost

fix(litellm): gate cost_log_tags on agent capability, not route (fixes non-Claude crash)

fix(litellm): proxy-authoritative token buckets + all-priced gate + transactional join

fix(litellm): per-attempt cost-log scoping + single run-id accessor + no-match warning

fix(litellm): sanitize cost headers, reject non-finite cost, drop debug scaffolding

docs(litellm): document LITELLM_COST_LOG wiring + correct the reconciliation-cost contract

refactor(litellm): stream the cost log + de-duplicate the OpenRouter config comment

refactor(litellm): cut per-message distribution; turn-level join + per-call audit record

docs(litellm): correct the cost contract after cutting per-message distribution

feat(evalboard): per-call cost/cache table from provider_call_costs (replaces inline)

Commits on Jul 31, 2026

feat(evalboard): make each turn's provider-call table a collapsed dropdown

fix(litellm): make the orphaned-spend warning actually fire

refactor(litellm): drop the provider field/column — unavailable on the streaming path

feat(docker): bind-mount the LiteLLM cost log so --driver docker joins actual cost

Merge branch 'main' into feat/openweight-actual-cost