I am new to this group and just starting with Beancount (after reading the docs for 2+ years!).
After looking at the current state, I decided to jump right in with v3. My main focus at the moment is entry. I have about 20+ bank accounts and about 200 transactions a month, many of which are automated or recurring, so it really makes no sense to enter them manually. Although I understand the idea behind the manual transaction entry philosophy.
Anyway, I am in Germany and there is a fairly widely used OSS software for collecting data from banks: Hibiscus (
willuhn.de/products/hibiscus/). Hibiscus supports 90% of my bank APIs and I use it to retrieve transactions (but also to make transactions, change recurring payments, etc.). A very nice tool that I plan to continue using.
My fallback strategy for beancount ingestion would be the following:
1. Hibiscus H2 Database
2. OFXget
3. Manual CSV Download
(4. PDF parsing)
I am starting with the beangulp ingestion from the H2 Database.
My question: I can use jaydebeapi to directly connect to the Hibiscus H2 DB from Python. I think this would be a single importer in beangulp, correct? I don't need bean-identify, because I know that I want to connect to the H2 DB already. Given the Beangulp Example (
https://github.com/beancount/beangulp/tree/master/examples), where would I best add the hook for the DB-connect - perhaps in process_extracted_entries()?
Also, I do not add categories (re: beancount accounts) to transactions in Hibiscus. This would be a manual step of assigning categories. I am wondering if this should be an ad hoc step once the data is in the beancount ledger, or if I should do it as part of the beangulp ingestion process. Are there any conventions for preferring one over the other?
Many thanks, Alex