📈 Reports & Caching

Reports aggregate expense data; caching keeps the hot paths fast without Mongo round-trips.

📊 Reports

Report endpoints live under interfaces/report/ and run over the expenses via the report domain models in domain/report/.

Aggregate by tag, entity, category, or date range.
Output shapes are defined in domain report models and serialized by the interface DTOs.
Shared with the CLI report helpers when driven locally.

Typical flow: GET /reports/{kind}?from=...&to=...&tag=... returns a grouped breakdown (count + total per bucket).

Repos in infra/repos/ are generic mongo-ops repos wrapped per model. Key behaviors:

search(filter={...}, limit=...) — filter-first reads, no magic.
In-memory LRU cache — the default read path; writes call _invalidate_cache.
Optional Redis mirror — infra/cache/redis.py with TTL-aware caching.
x-fk cross-collection population — referenced ids (e.g. account_id on expenses) are populated by explicit filter lookups against the target repo, not Mongo joins.

infra/http_cache/ caches external HTTP responses:

Config	Meaning
`HTTP_CACHE_TTL`	Cache time-to-live for responses
`HTTP_CACHE_MAX_ENTRIES`	Eviction cap for the cache store
`REDIS_URL`	When set, cache lives in Redis; otherwise in-memory

Used by fetchers and enrichment lookups to avoid hammering upstreams during a full-statement import.

Env var	Purpose
`MONGO_URI` / `MONGO_DB`	MongoDB connection + database
`REDIS_URL`	Redis (cache mirror, optional)
`HTTP_CACHE_TTL` / `HTTP_CACHE_MAX_ENTRIES`	HTTP cache behavior
`LLAMA_CLOUD_API_KEY`	LlamaParse (llama family)
`EMAIL_HOST` / `EMAIL_USER` / `EMAIL_PASS`	IMAP ingestion
`AUTH_SERVER`	External auth/service integration
`WORKER_*`	Worker polling knobs (see fetch-request workflow)
`DEBUG`	Verbose logging