mirror of
https://github.com/tiennm99/miti99bot.git
synced 2026-09-13 22:19:08 +00:00
tiennm99
1fc425c55f
Removes six modules (loldle-ability/emoji/quote/splash, semantle, doantu) and prunes the framework deps that were only there to serve them: - ai.Embedder + Client.Embed + embeddingModel const (semantle only) - Deps.Embedder + BuildOptions.Embedder - Deps.Env + Build(env) param + ModuleEnv config field + PHOW2SIM allowlist (doantu only) - internal/champname package (loldle now owns its lookup helpers directly) - template.yaml: Phow2simAPIURL parameter + PHOW2SIM_API_URL Lambda env Active catalog: util, misc, wordle, loldle, lolschedule, twentyq, trading. go build / vet / test all pass.
23 lines
946 B
Go
23 lines
946 B
Go
// Package ai wraps Google's genai SDK with a small, mockable surface for the
|
|
// twentyq module. The package owns:
|
|
//
|
|
// - Chatter interface — what modules consume
|
|
// - Client struct — production implementation backed by genai
|
|
// - per-user rate limiter — defends the shared 1500-RPD Gemini free tier
|
|
//
|
|
// Modules accept the interface (not *Client) so unit tests can pass fakes
|
|
// without spinning up a real Gemini client. Production wiring in cmd/server
|
|
// passes the *Client (which satisfies the interface) into Deps.
|
|
package ai
|
|
|
|
import "context"
|
|
|
|
// Chatter produces a single text completion from a system + user prompt
|
|
// pair. Used by the twentyq module's judge + round-start calls.
|
|
//
|
|
// On rate-limit (HTTP 429) the implementation should return ErrRateLimited
|
|
// so callers can show a user-friendly retry message.
|
|
type Chatter interface {
|
|
Generate(ctx context.Context, system, user string) (string, error)
|
|
}
|