Function createChartReviewChatModel

createChartReviewChatModel(
    settings: {
        enableCalendarAwareness?: boolean;
        enableEpisodePatientBlock?: boolean;
        enableExpandedStatuses?: boolean;
        enableIntentAwarePlanning?: boolean;
        enableMetadataAwareGate?: boolean;
        enablePerQuestionInstructions?: boolean;
        enableQueryHydration?: boolean;
        enableReviewCriteriaInDocumentSelection?: boolean;
        enableRunConstantPrefix?: boolean;
        enableStatusReconciliation?: boolean;
        enableSubClaimFanOut?: boolean;
        enableSummaryTableShortcut?: boolean;
        enableThinRetrievalBailout?: boolean;
        enableTimeFilteringInChartReview?: boolean;
        enableUnlockedRetrieval?: boolean;
        enableV3PassFailRunner?: boolean;
        maxTokens: number;
    },
    opts?: { promptCacheKey?: string },
): ChatOpenAI
Constructs the ChatOpenAI instance that powers a chart-review run.

The values mirror v1's circlehealthai-web/app/api/chart-review/run-v2/route.ts verbatim (gpt-4.1, temperature 0.1, no streaming, 3 retries). maxTokens is sourced from the resolved ChartReviewSettings so LD overrides flow through.

Lives here (rather than in the trpc handler) so the @langchain/openai dependency stays confined to @repo/legacy.

promptCacheKey routes same-prefix traffic to one machine so the automatic prompt cache is realized at volume. Pass it only from single-purpose call sites: the shared run-wide instance serves several prompts, and one key across different prefixes just dilutes the routing.
Parameters
- settings: {
      enableCalendarAwareness?: boolean;
      enableEpisodePatientBlock?: boolean;
      enableExpandedStatuses?: boolean;
      enableIntentAwarePlanning?: boolean;
      enableMetadataAwareGate?: boolean;
      enablePerQuestionInstructions?: boolean;
      enableQueryHydration?: boolean;
      enableReviewCriteriaInDocumentSelection?: boolean;
      enableRunConstantPrefix?: boolean;
      enableStatusReconciliation?: boolean;
      enableSubClaimFanOut?: boolean;
      enableSummaryTableShortcut?: boolean;
      enableThinRetrievalBailout?: boolean;
      enableTimeFilteringInChartReview?: boolean;
      enableUnlockedRetrieval?: boolean;
      enableV3PassFailRunner?: boolean;
      maxTokens: number;
  }
  - OptionalenableCalendarAwareness?: boolean
  - OptionalenableEpisodePatientBlock?: boolean
  - OptionalenableExpandedStatuses?: boolean
  - OptionalenableIntentAwarePlanning?: boolean
  - OptionalenableMetadataAwareGate?: boolean
  - OptionalenablePerQuestionInstructions?: boolean
  - OptionalenableQueryHydration?: boolean
  - OptionalenableReviewCriteriaInDocumentSelection?: boolean
  - OptionalenableRunConstantPrefix?: boolean
    Only read when enableV3PassFailRunner is on — it configures the v3 runner, not v2.
  - OptionalenableStatusReconciliation?: boolean
  - OptionalenableSubClaimFanOut?: boolean
    Only read when enableV3PassFailRunner is on — it configures the v3 runner, not v2.
  - OptionalenableSummaryTableShortcut?: boolean
  - OptionalenableThinRetrievalBailout?: boolean
  - OptionalenableTimeFilteringInChartReview?: boolean
  - OptionalenableUnlockedRetrieval?: boolean
  - OptionalenableV3PassFailRunner?: boolean
  - maxTokens: number
- Optionalopts: { promptCacheKey?: string }
Returns ChatOpenAI
- Defined in packages/legacy/src/chart-review/run-v2/createChatModel.ts:19