The platform now runs on Groq by default, through the OpenAI-compatible chat-completions shape. That shape is not one vendor — Gemini, OpenRouter, Together, vLLM and a local Ollama serve it too — so moving again stays configuration rather than code. Two things in the deleted file were not Anthropic's and would have gone with it silently: withRetry / MaxAttempts / retryBackoff were defined in anthropic.go and CALLED BY openai.go. Deleting the file wholesale would have removed the retry policy of the provider that survived, and nothing in openai.go mentions it, so the loss would have been invisible until the next 429. The policy is a property of this platform's runs, not of a vendor's API; it now lives in retry.go where no provider can carry it off. StreamComplete had the same problem and moves to gateway.go, beside the Streamer interface whose comment already referenced it. Three stale-configuration failures are now refused at startup instead of being ignored. Each was verified firing through the real config.Load(): MODEL_PROVIDER=anthropic — named separately from every other wrong value because it used to be correct. Ignoring it gives a stack that believes it is on Claude while every run goes to Groq and is billed there. ANTHROPIC_API_KEY set while MODEL_API_KEY is empty. Ignoring a key an operator did set is the worst version of this: they fail every run on a missing credential they are looking straight at. A leftover claude-* model id, naming the tier that carries it. This is the check the previous commit's error-detail work was diagnosing: such an id is accepted by this process, rejected by the provider, and 400s on EVERY run. "A model is wrong" does not say which of three lines to edit. Defaults ship as a matched pair. defaultBaseURL and the three tier ids are one decision, not four: an id is only meaningful against the service that serves it, and a Groq id on an OpenAI base URL is the same failure from the other side. The tiers also stop being one model — a tier whose cost does not differ is a distinction that buys nothing. Verified end to end against a stub of the wire, driving the real wiring (config.Load in production mode, gateway.New, StreamComplete): streamed deltas, tool-call decoding, the loopback credential exemption, and usage totalling 150 rather than 190 — the cached-prefix subtraction still holds. gofmt clean, go vet clean, 14/14 non-DB packages pass. httpserver still needs a reachable database. NOT verified: the I7 planted-injection eval. Removing this path removed the only model whose refusal behaviour had been measured against it, so the new default is unproven there until `make eval-live` runs with a real key. The Groq model ids should also be confirmed against Groq's current lineup. Flagged in CLAUDE.md §12 and docs/handover.md. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PJvibeSc1JYXjatankqM1g
165 lines
5.2 KiB
Go
165 lines
5.2 KiB
Go
package runtime_test
|
|
|
|
import (
|
|
"context"
|
|
"encoding/json"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/krow/krow-backend/go-api/internal/runtime"
|
|
"github.com/krow/krow-backend/go-api/internal/testutil"
|
|
)
|
|
|
|
// trajectory builds a saveable run for the given harness org.
|
|
func trajectory(orgID, runID string) *runtime.Trajectory {
|
|
started := time.Now().Add(-2 * time.Second).UTC()
|
|
return &runtime.Trajectory{
|
|
RunID: runID,
|
|
OrgID: orgID,
|
|
AgentID: "activity-agent",
|
|
AgentVersion: 3,
|
|
Tier: "balanced",
|
|
Model: "llama-3.3-70b-versatile",
|
|
StartedAt: started,
|
|
EndedAt: started.Add(1200 * time.Millisecond),
|
|
Termination: runtime.TerminationCompleted,
|
|
Entries: []runtime.Entry{
|
|
{Seq: 1, At: started, Kind: runtime.EntryMessage, Role: "user", Text: "what happened?"},
|
|
{Seq: 2, At: started, Kind: runtime.EntryBudget, Budget: &runtime.Snapshot{StepsLeft: 8, TokensLeft: 120000}},
|
|
{Seq: 3, At: started, Kind: runtime.EntryMessage, Role: "assistant", Text: "Twelve events."},
|
|
},
|
|
Usage: runtime.RunUsage{InputTokens: 900, OutputTokens: 120, TotalTokens: 1020, ModelCalls: 1},
|
|
}
|
|
}
|
|
|
|
func TestPostgresSinkSavesAndReadsBack(t *testing.T) {
|
|
h := testutil.New(t)
|
|
ctx := context.Background()
|
|
sink := runtime.NewPostgresSink(h.Pool)
|
|
|
|
traj := trajectory(h.OrgID, "run_store_basic")
|
|
if err := sink.Save(ctx, traj); err != nil {
|
|
t.Fatalf("save: %v", err)
|
|
}
|
|
|
|
var (
|
|
agentID, tier, model, termination string
|
|
version, modelCalls int
|
|
total int64
|
|
entries []byte
|
|
)
|
|
err := h.Pool.QueryRow(ctx, `
|
|
SELECT agent_id, agent_version, tier, model, termination, total_tokens, model_calls, entries
|
|
FROM agent_runs WHERE run_id = $1`, traj.RunID,
|
|
).Scan(&agentID, &version, &tier, &model, &termination, &total, &modelCalls, &entries)
|
|
if err != nil {
|
|
t.Fatalf("read back: %v", err)
|
|
}
|
|
|
|
if agentID != "activity-agent" || version != 3 {
|
|
t.Errorf("stored %s v%d, want activity-agent v3", agentID, version)
|
|
}
|
|
if termination != string(runtime.TerminationCompleted) {
|
|
t.Errorf("termination = %q, want Completed", termination)
|
|
}
|
|
// Both the tier asked for and the model that answered, so a trajectory read
|
|
// a year later does not require knowing that week's routing.
|
|
if tier != "balanced" || model != "llama-3.3-70b-versatile" {
|
|
t.Errorf("tier/model = %q/%q, want balanced/llama-3.3-70b-versatile", tier, model)
|
|
}
|
|
if total != 1020 || modelCalls != 1 {
|
|
t.Errorf("usage = %d tokens over %d calls, want 1020/1", total, modelCalls)
|
|
}
|
|
|
|
var round []runtime.Entry
|
|
if err := json.Unmarshal(entries, &round); err != nil {
|
|
t.Fatalf("entries did not round-trip: %v", err)
|
|
}
|
|
if len(round) != 3 || round[0].Role != "user" || round[2].Text != "Twelve events." {
|
|
t.Errorf("entries round-tripped as %+v", round)
|
|
}
|
|
}
|
|
|
|
func TestPostgresSinkIsIdempotentPerRun(t *testing.T) {
|
|
// A run id is generated once and written once. A second save is a retry of
|
|
// one that already landed — failing it would turn a harmless duplicate
|
|
// into a reported error on a run that succeeded.
|
|
h := testutil.New(t)
|
|
ctx := context.Background()
|
|
sink := runtime.NewPostgresSink(h.Pool)
|
|
|
|
traj := trajectory(h.OrgID, "run_store_twice")
|
|
if err := sink.Save(ctx, traj); err != nil {
|
|
t.Fatalf("first save: %v", err)
|
|
}
|
|
if err := sink.Save(ctx, traj); err != nil {
|
|
t.Fatalf("second save should be a no-op, got: %v", err)
|
|
}
|
|
|
|
var n int
|
|
if err := h.Pool.QueryRow(ctx,
|
|
`SELECT count(*) FROM agent_runs WHERE run_id = $1`, traj.RunID).Scan(&n); err != nil {
|
|
t.Fatalf("count: %v", err)
|
|
}
|
|
if n != 1 {
|
|
t.Errorf("%d rows for one run id, want 1", n)
|
|
}
|
|
}
|
|
|
|
func TestPostgresSinkRefusesRunsItCannotAttribute(t *testing.T) {
|
|
h := testutil.New(t)
|
|
ctx := context.Background()
|
|
sink := runtime.NewPostgresSink(h.Pool)
|
|
|
|
cases := map[string]*runtime.Trajectory{
|
|
"no run id": func() *runtime.Trajectory {
|
|
tr := trajectory(h.OrgID, "")
|
|
return tr
|
|
}(),
|
|
// I5: tenancy is not optional. Caught here so the failure names the
|
|
// cause rather than surfacing as a NOT NULL violation.
|
|
"no org": func() *runtime.Trajectory {
|
|
tr := trajectory("", "run_no_org")
|
|
return tr
|
|
}(),
|
|
// An invented termination must never reach the column that evals and
|
|
// dashboards group by.
|
|
"invented termination": func() *runtime.Trajectory {
|
|
tr := trajectory(h.OrgID, "run_bad_term")
|
|
tr.Termination = "Finished"
|
|
return tr
|
|
}(),
|
|
}
|
|
|
|
for name, tr := range cases {
|
|
if err := sink.Save(ctx, tr); err == nil {
|
|
t.Errorf("%s: save should have been refused", name)
|
|
}
|
|
}
|
|
}
|
|
|
|
func TestPostgresSinkStoresAnEmptyTrajectory(t *testing.T) {
|
|
// A run that recorded nothing is still a run worth keeping, and a nil
|
|
// slice marshals to "null", which the jsonb_typeof CHECK refuses.
|
|
h := testutil.New(t)
|
|
ctx := context.Background()
|
|
sink := runtime.NewPostgresSink(h.Pool)
|
|
|
|
traj := trajectory(h.OrgID, "run_store_empty")
|
|
traj.Entries = nil
|
|
traj.Termination = runtime.TerminationBudgetExceeded
|
|
|
|
if err := sink.Save(ctx, traj); err != nil {
|
|
t.Fatalf("save: %v", err)
|
|
}
|
|
|
|
var kind string
|
|
if err := h.Pool.QueryRow(ctx,
|
|
`SELECT jsonb_typeof(entries) FROM agent_runs WHERE run_id = $1`, traj.RunID).Scan(&kind); err != nil {
|
|
t.Fatalf("read back: %v", err)
|
|
}
|
|
if kind != "array" {
|
|
t.Errorf("entries stored as %q, want array", kind)
|
|
}
|
|
}
|