[e2e] drive the provider matrix with a long-form completion

Switch the matrix prompt to 'explain GitHub workflow in 1000 words' and
raise the harness bodies' max_tokens from 64 to 2048 so every provider
produces a realistic ~1k-word completion. Cost validation (API recompute
and raw-SQL store audit) then runs on token counts of the same order as
real traffic instead of a one-word pong.
This commit is contained in:
Maycon Santos
2026-07-25 05:33:14 +00:00
parent f9f7995ebe
commit 858d492109
2 changed files with 13 additions and 6 deletions

View File

@@ -429,6 +429,13 @@ func TestProvidersMatrix(t *testing.T) {
// session id and confirm the marker propagated end-to-end.
sessionID := "e2e-session-" + pc.name
// A long-form prompt so every provider produces a realistic
// (~1k-word) completion — cost validation then runs on token
// counts of the same order as real traffic instead of a
// one-word pong. max_tokens in the harness bodies is sized
// (2048) to let the full answer through.
const matrixPrompt = "explain GitHub workflow in 1000 words"
// Retry briefly to absorb tunnel/DNS jitter on the first call.
var code int
var body string
@@ -439,11 +446,11 @@ func TestProvidersMatrix(t *testing.T) {
var cerr error
switch pc.kind {
case harness.WireVertex:
c, b, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, pc.project, pc.region, pc.model, "Reply with exactly: pong", sessionID)
c, b, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, pc.project, pc.region, pc.model, matrixPrompt, sessionID)
case harness.WireBedrock:
c, b, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, pc.model, "Reply with exactly: pong", sessionID)
c, b, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, pc.model, matrixPrompt, sessionID)
default:
c, b, cerr = cl.ChatPrefixed(ctx, settings.Endpoint, proxyIP, pc.pathPrefix, pc.kind, pc.model, "Reply with exactly: pong", sessionID)
c, b, cerr = cl.ChatPrefixed(ctx, settings.Endpoint, proxyIP, pc.pathPrefix, pc.kind, pc.model, matrixPrompt, sessionID)
}
if cerr == nil {
code, body = c, b

View File

@@ -256,7 +256,7 @@ func (cl *Client) ChatPrefixed(ctx context.Context, endpoint, proxyIP, pathPrefi
case WireMessages:
path = "/v1/messages"
headers = []string{"anthropic-version: 2023-06-01"}
body = fmt.Sprintf(`{"model":%q,"max_tokens":64,"messages":[{"role":"user","content":%q}]}`, model, prompt)
body = fmt.Sprintf(`{"model":%q,"max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, model, prompt)
default:
path = "/v1/chat/completions"
body = fmt.Sprintf(`{"model":%q,"messages":[{"role":"user","content":%q}]}`, model, prompt)
@@ -271,7 +271,7 @@ func (cl *Client) ChatPrefixed(ctx context.Context, endpoint, proxyIP, pathPrefi
// is sent as the universal x-session-id header the proxy records.
func (cl *Client) Vertex(ctx context.Context, endpoint, proxyIP, project, region, model, prompt, sessionID string) (int, string, error) {
path := fmt.Sprintf("/v1/projects/%s/locations/%s/publishers/anthropic/models/%s:rawPredict", project, region, model)
body := fmt.Sprintf(`{"anthropic_version":"vertex-2023-10-16","max_tokens":64,"messages":[{"role":"user","content":%q}]}`, prompt)
body := fmt.Sprintf(`{"anthropic_version":"vertex-2023-10-16","max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, prompt)
return cl.post(ctx, endpoint, proxyIP, path, body, withSessionID(nil, sessionID))
}
@@ -282,7 +282,7 @@ func (cl *Client) Vertex(ctx context.Context, endpoint, proxyIP, project, region
// header the proxy records.
func (cl *Client) Bedrock(ctx context.Context, endpoint, proxyIP, model, prompt, sessionID string) (int, string, error) {
path := "/model/" + model + "/invoke"
body := fmt.Sprintf(`{"anthropic_version":"bedrock-2023-05-31","max_tokens":64,"messages":[{"role":"user","content":%q}]}`, prompt)
body := fmt.Sprintf(`{"anthropic_version":"bedrock-2023-05-31","max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, prompt)
return cl.post(ctx, endpoint, proxyIP, path, body, withSessionID(nil, sessionID))
}