fix: add LLM_API_URL and LLM_MODEL to portfolio deployment
Problem: Portfolio pod had OAuth credentials (AUTHENTIK_CLIENT_ID/SECRET) but was missing LLM_API_URL and LLM_MODEL environment variables. Chat API could fetch tokens but had no endpoint to call. Solution: Inject LLM configuration via k8s deployment env vars: - LLM_API_URL: https://api.riotpiao.com/v1/chat/completions - LLM_MODEL: qwen2.5:3b-instruct Testing: - Token fetch: ✓ portfolio-agent OAuth works - DNS: ✓ api.riotpiao.com resolves (CoreDNS rewrite) - Auth: ⚠️ JWT issuer mismatch (portfolio-agent vs api-gw) — separate issue Impact: Chat API endpoint can now attempt LLM calls with valid auth
This commit is contained in:
@@ -42,6 +42,11 @@ spec:
|
||||
secretKeyRef:
|
||||
name: portfolio-agent-oidc
|
||||
key: TOKEN_URL
|
||||
# LLM API configuration
|
||||
- name: LLM_API_URL
|
||||
value: "https://api.riotpiao.com/v1/chat/completions"
|
||||
- name: LLM_MODEL
|
||||
value: "qwen2.5:3b-instruct"
|
||||
ports:
|
||||
- name: http
|
||||
containerPort: 3000
|
||||
|
||||
Reference in New Issue
Block a user