Repository Issue Activity (beta)

#5788 [Bug]: Long-context tier starts one token early — getPricingTier is inclusive, Google and OpenAI document their thresholds as exclusive

#5805 openAIUsageProcessor: accepted_prediction_tokens added (not subtracted) into output — cost overstated for Predicted Outputs

#5811 [Feature]: OrcaRouter provider support for Helicone's AI Gateway

#5809 Cost registry: how should a provider whose model ids are market instruments be represented?

#5746 Experiment evaluator run can return success after an evaluator error

#5753 Document and CI - Validate Valkey Support for Jawn

#5612 Proposal: ship WFGY 16-problem RAG failure labels as an optional Helicone tagging scheme

#5795 OpenAI-compat: gateway /v1 plus origin-only Helicone-Target-URL, not oai.helicone.ai

#5610 MCP: complex object parameters (filter, sort) fail with clients that serialize them as JSON strings

#5794 Adding a gateway to the cost registry: which shape do you want?

#5550 Issue on docs

#5597 Security: Cross-Organization Provider Key IDOR in Vault Endpoint

#5790 docs.helicone.ai llms.txt: two dead internal links (/features/prompts/generate, /features/prompts/editor → 404)

#5787 Optional EvalPort interop for datasets and experiment scores

#5784 How to privately report a security vulnerability? (PVR not enabled; security@ bounced)

#5782 dbProviderToProvider silently drops xai/perplexity/mistral/helicone provider keys — 4 of 21 providers missing

#5780 cacheSettings regex matches x-s-maxage (and other vendor extensions) as s-maxage — no word boundary before directive name

#5775 secureCache.getFromKVCacheOnly and getAndStoreInCache ignore the expirationTtl argument — caches effectively 1-minute not 12-hour

#5771 worker: 3 Promise.race + setTimeout sites never clearTimeout on early resolution — pending timer leaks

#5768 Helicone-Cache-Ignore-Keys is documented as comma-separated but parsed as wrapped JSON — single-element array, broken cache-key ignore

#5561 [Bug]: AI Gateway (ai-gateway.helicone.ai) ignores provided Gemini API key, uses free-tier keys instead

#5767 getThresholdValueFunction ignores bedrock/helicone providers, so tiered Claude never applies the >200k tier there

#5766 Cache-write cost ignores long-context pricing tier (uses base tier, ~2x undercharge on >200k Anthropic requests)

#5763 [Feature]: Governance Enforcement Layer Alongside Helicone Observability

#5645 Integration: Cryptographic signing for LLM logs with asqav