ProjectTech4DevAI / ProjectTech4DevAI/kaapi-backend
API Management: Standardizing key usage
Dieses Issue hat noch niemand übernommen.
- Vorherrschende Sprache
- Python
- Sterne
- 18
- Forks
- 10
- Ø Merge
- 2 T. 20 Std.
- Gemergte PRs (30 T.)
- 14
Beschreibung
Is your feature request related to a problem?
OpenAI API keys are dispersed across individual developer accounts, leading to unreliable cost attribution and making it difficult to track usage. The absence of a naming and ownership convention results in many keys that are hard to identify, further complicating management.
Describe the solution you'd like
-
Inventory
- Export all active API keys across Kaapi (id, project, creation date, creator, last used, spend over the last 6 months).
- Map each key to a use case and a current owner.
- Flag keys with no identifiable owner or no usage in the last 90 days.
-
Clean up
- Revoke unused and orphaned keys.
- Create replacements for in-use keys under the shared account, rotate the consuming services, then revoke originals.
- Verify nothing breaks post-rotation (staging first).
-
Standardise
- Document key creation convention from the shared account, not personal accounts.
- Use projects/service accounts for workload separation.
- Establish a naming scheme (e.g.,
<service>-<env>-<purpose>). - Define ownership for key creation and revocation.
- Add the convention to onboarding docs.
Original issue
Context
OpenAI API keys are currently spread across individual developer accounts rather
than the shared organisational account. Most keys appear to have been created
ad hoc by whoever needed one at the time, across a range of use cases (evals,
guardrails, feature work), so spend concentrates under a few personal accounts
and the shared account shows almost no usage.
There is also no naming or ownership convention, so it is hard to tell what any
given key is used for or whether it is still live. Several keys have been active
for a long time with no clear owner. This makes cost attribution unreliable and
the platform dashboard hard to read — pulling a six-month spend total currently
requires querying the usage API manually.
Scope
ONLY TOUCH KEYS USED BY KAAPI. DO NOT TOUCH OTHER KEYS (EG. GLIFIC BOTS, OTHER ORG KEYS)
1. Inventory
- Export all active API keys across Kaapi (id, project, creation date, creator, last used, spend over the last 6 months)
- Map each key to a use case and a current owner
- Flag keys with no identifiable owner or no usage in the last 90 days
2. Clean up
- Revoke unused and orphaned keys
- For keys still in use, create replacements under the shared account, rotate the consuming services over, then revoke the originals
- Verify nothing breaks post-rotation (staging first wherever a key backs a production path)
3. Standardise
- Document the convention: keys for regular operational use are created from the shared account, never from personal accounts
- Use projects / service accounts to separate workloads instead of separate personal keys, so usage is attributable by default
- Agree a naming scheme, e.g.
<service>-<env>-<purpose> - Define who owns key creation and revocation, and where the register of keys lives
- Add the convention to onboarding docs
Acceptance criteria
- No active keys remain under personal accounts for operational workloads
- Every active key has a documented owner, purpose, and conventional name
- Convention is written down and linked from onboarding
Notes
- Rotation needs coordination with whoever owns each consuming service; worth
identifying those owners during the inventory step rather than at rotation time. - Personal keys used purely for local experimentation can stay out of scope, but
should be confirmed as non-production before being left alone.
Beitragsleitfaden
Erste Schritte
- Lies das ganze Issue und danach den Beitragsleitfaden des Projekts.
- Schreib ins Issue, dass du es übernimmst — das erspart doppelte Arbeit.
- Forke das Repository und arbeite in einem Branch.
- Öffne einen Pull Request, der die Issue-Nummer nennt.
Rechercherichtung
Es werden keine Repository-Dateien, Tests oder Code-Einstiegspunkte genannt. Beginne damit, dem Umfang der Bestandsaufnahme in diesem Issue und der darin genannten Usage API zu folgen, und ordne dann die Kaapi-eigenen Keys ihren Verantwortlichen und Services zu, bevor du die Rotation planst. Als erledigt gilt die Aufgabe, wenn sich die operativen Keys unter dem gemeinsamen Account befinden, jeder einen Verantwortlichen, einen Zweck und einen konventionellen Namen hat und die Konvention aus der Onboarding-Dokumentation verlinkt ist.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Bewertung
- Bereich
- documentation, security
- Issue-Typ
- Feature
- Schwierigkeit
- 5/5
- Geschätzter Aufwand
- Über eine Woche
- Aktivitätsstatus
- Ruhig
- Klarheit
- Größtenteils klar
- Anfängerfreundlichkeit
- 30/100