Greptile Summary

This PR adds multi-instance support to five provider integrations (Grafana, AWS, Datadog, Honeycomb, Coralogix) through a new v2 store schema, *_INSTANCES JSON env vars, and selector helpers. Legacy single-instance configs are fully preserved via backward-compat flat keys and auto-migration of v1 store records on load.

The three P1 regressions from PR #527 are explicitly addressed (AWS role_arn path, env data loss via merge_integrations_by_service, filtered-get leak), the propagation condition in resolve_effective_integrations now correctly mirrors classify_integrations's publication condition, _STRUCTURAL_RECORD_FIELDS is no longer duplicated (catalog.py imports from store.py), and the Grafana hint warning is in place. One P2 remains: tag values from _parse_instances_env are stored verbatim without lowercasing, while the selector helpers normalize only the search side — silent mismatches for any env-sourced instance whose JSON tags contain uppercase characters.

Confidence Score: 5/5

Safe to merge — all previously flagged P1 issues are resolved and the one remaining finding is a P2 edge case affecting only uppercase tag values in env-sourced instances.

All three PR #527 P1 bugs are fixed with named regression tests. Previous review threads about the propagation condition mismatch, _STRUCTURAL_RECORD_FIELDS duplication, and the Grafana hint silent fallback are all addressed in this revision. The only new finding (tag case normalization for env-sourced instances) is a P2 that only bites users who intentionally supply uppercase tag values in *_INSTANCES JSON, which is unlikely given the documentation examples all use lowercase.

app/integrations/catalog.py — tag normalization gap in _parse_instances_env (P2).

Important Files Changed

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Env Vars\n*_INSTANCES JSON\nor legacy single vars] --> B[load_env_integrations]
    C[~/.tracer/integrations.json\nv1 or v2] --> D[load_integrations\nauto-migrate v1→v2]
    B --> E[merge_integrations_by_service\nstore wins over env]
    D --> E
    E --> F[classify_integrations\nfor each instance in each record]
    F --> G[resolved service\nflat default-instance config]
    F --> H[_all_service_instances\nlist of name+tags+config\nonly when multi or non-default name]
    G --> I[selectors.py\nget_default_instance / get_instance_by_name\nget_instances_by_tag / select_instance]
    H --> I
    I --> J[detect_sources.py\ngrafana_instance hint lookup]
    F --> K[resolve_effective_integrations\nEffectiveIntegrations model dump\nwith instances propagated when present]

This is a comment left during a code review.
Path: app/integrations/catalog.py
Line: 593-594

Comment:
**Tag values not normalized to lowercase for env-sourced instances**

`_parse_instances_env` stores tag values verbatim from the JSON input. Both `selectors.get_instances_by_tag` and `selectors.select_instance` normalize the *search* value to lowercase (e.g., `target_value = (value or "").strip().lower()`) but compare directly against the stored, un-normalized tag value. A user who writes `"tags": {"env": "Prod"}` in `GRAFANA_INSTANCES` JSON would get `{"env": "Prod"}` stored, but any tag-based selection call like `get_instances_by_tag(..., "env", "prod")` silently returns nothing because `"Prod" != "prod"`.

The `IntegrationInstance._normalize_tags` validator in `models.py` already enforces lowercase tag values, but it is never invoked in this env-parsing path. Adding `.lower()` here would make env-sourced instances behave the same as store-sourced ones.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (5): Last reviewed commit: "merge: resolve catalog.py conflict with ..." | Re-trigger Greptile}

@Devesh36 please again review.
Thanks

Yeah, looks good. Really appreciate the contribution. @hamzzaaamalik