lars/finance-duck
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Implement classification redesign

Browse Source
This commit is contained in:
Lars Nolden
2026-09-11 22:46:17 +02:00
parent cc43a2f9a7
commit 87f052a3ea
23 changed files with 1602 additions and 296 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+6 -2
View File
@@ -2,7 +2,7 @@
A self-hosted personal finance dashboard with a **Go backend**, **React frontend**, and **DuckDB analytics**. Human-readable `.finance` journals are the source of truth; DuckDB is a disposable index.
Imported bank facts are separate from editable merchant, category, and tag classifications. N26, ING, Kontist, and Scalable Capital CSV imports and bank synchronization work without AI. An investment account tracks positions by ISIN alongside its cash, and reconciles both against your broker's own figures. Optional OpenRouter enrichment uses restrictive provider routing and omits amounts by default, and can map the columns of an unrecognized CSV layout from a redacted sample.
Imported bank facts are separate from editable merchant, category, and tag classifications. N26, ING, Kontist, and Scalable Capital CSV imports and bank synchronization work without AI. An investment account tracks positions by ISIN alongside its cash, and reconciles both against your broker's own figures. Optional OpenRouter enrichment sends the transaction date, signed amount, currency, merchant/counterparty text, and a complete registry of editable classification choices through restrictive private routing.
> **There is no application login.** Keep Finance Duck behind your VPN. The default service and Docker Compose port bindings are loopback-only. Setting a hostname does not provide authentication or firewall protection.
@@ -401,7 +401,11 @@ Bank synchronization and recognized N26, ING, and Kontist CSV imports do **not**
**Classify newly imported transactions with AI** under **Classification preferences** controls whether importing contacts the provider at all. It covers CSV imports and bank synchronization, is on by default, and is stored as `classify_on_import` in `config.toml`. With it off, no import makes a provider request: enabled merchant rules still classify, and everything else arrives unclassified and editable without a failure that would suggest the provider was unreachable. **AI classification → Analyse** still works on demand, so you can review a batch deliberately instead of on every import.
Every AI classification requests `provider.data_collection = "deny"`, `provider.zdr = true`, and `provider.require_parameters = true`. Unsupported private routing fails rather than falling back to a less restrictive provider. Amount sharing is off by default. Keep OpenRouter account prompt logging disabled as well. Automatic redaction minimizes data; it is not a guarantee that arbitrary transaction prose is anonymous.
AI classification sends only identifier-redacted text: the transaction's own IDs, account identifiers, payment references, and configured private names are removed, while merchant and counterparty text remains available for recognition. Classification responses carry `high`, `medium`, or `low` confidence; low-confidence results retain the merchant and tags but use the kind-appropriate unclassified category and appear in **Transactions → Needs review**.
From **Categories**, **Propose taxonomy** samples up to 300 redacted transactions, grouped so recurring counterparties are represented without sending raw identifiers. The proposal can suggest categories, tags, and merchants with hints and evidence. Approve each item individually; applying it also creates any approved category parents required by the hierarchy. Existing registry entries and transaction facts are never overwritten.
Every AI classification requests `provider.data_collection = "deny"`, `provider.zdr = true`, and `provider.require_parameters = true`. Unsupported private routing fails rather than falling back to a less restrictive provider. Amount, date, and currency are always included; identifier-only redaction removes account and transaction identifiers, payment references, and configured private names but does not remove merchant or counterparty text. Keep OpenRouter account prompt logging disabled as well.
Classification spaces request starts by at least **three seconds**, including successful requests, rather than sending a burst between 429s. This is a conservative application policy, not a published quota for every model. On HTTP 429, backoff starts at **15 seconds** and increases across consecutive failures; `Retry-After` seconds or HTTP dates can extend the wait. Successful retries retain the learned spacing (up to **30 seconds**) instead of immediately bursting again. Each operation makes at most **four attempts**, with at most **two minutes of automatic retry waiting**, preserving the same model, sanitized prompt, and privacy controls. Imports and previews share this pacing and cooldown. Long or exhausted limits leave records unclassified with a retry-time error; local merchant rules still work. After the cooldown, run **AI classification → Analyse** again for previously failed records—repeating a bank import does not reclassify existing transactions.