Skip to content

Strip believed-free models from pricing block to prevent cost attribution #108

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
BillJr99 merged 1 commit into from
Jun 19, 2026
Merged

Strip believed-free models from pricing block to prevent cost attribution #108

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 0 additions & 12 deletions llmproxy/providers.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1668,18 +1668,6 @@
"input_cost_per_token": 2.25e-06,
"output_cost_per_token": 2.75e-06
},
"cloudflare-workers/@cf/meta/llama-2-7b-chat-fp16": {
"input_cost_per_token": 1.923e-06,
"output_cost_per_token": 1.923e-06
},
"cloudflare-workers/@cf/meta/llama-2-7b-chat-int8": {
"input_cost_per_token": 1.923e-06,
"output_cost_per_token": 1.923e-06
},
"cloudflare-workers/@cf/mistral/mistral-7b-instruct-v0.1": {
"input_cost_per_token": 1.923e-06,
"output_cost_per_token": 1.923e-06
},
"cloudflare-workers/@hf/thebloke/codellama-7b-instruct-awq": {
"input_cost_per_token": 1.923e-06,
"output_cost_per_token": 1.923e-06
Expand Down
12 changes: 12 additions & 0 deletions scripts/update_free_models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -399,6 +399,12 @@ def _merge_pricing(sidecar: dict, updates: dict, litellm_ran: bool) -> bool:
could not be fetched this run, the existing block is kept as the baseline so a
transient outage never wipes pricing. Returns True if the block changed.
Network/parse failures are non-fatal — stale pricing beats aborting the run.

Believed-free models are always stripped from the final block. Some providers
(e.g. Cloudflare) have a paid-tier per-token price in the LiteLLM map even
though their free plan is quota-limited rather than zero-priced. Keeping those
entries causes compute_cost() to attribute a cost to every free-tier request,
which eventually lands the model in cost_observed_free_tier and breaks routing.
"""
existing = sidecar.get("pricing")
if not isinstance(existing, dict):
Expand All @@ -412,6 +418,12 @@ def _merge_pricing(sidecar: dict, updates: dict, litellm_ran: bool) -> bool:
baseline = existing
overrides = _collect_source_pricing(updates)
merged = {**baseline, **overrides}
# Strip any model that is currently believed_free — their cost accounting
# belongs in the free-tier path, not the paid pricing block.
free_models: set[str] = set()
for prov_data in sidecar.get("providers", {}).values():
free_models.update(m.lower() for m in prov_data.get("believed_free", []))
merged = {k: v for k, v in merged.items() if k not in free_models}
final = dict(sorted(merged.items()))
if final == existing:
return False
Expand Down
Loading