feat: lemonade provider serving validation - config contract and edge node mapping updates

- Migrate config contract files to archive (01_config_contract)
- Update edge config.go for new configuration structure
- Refactor node/mapper.go and mapper_test.go for field mapping
- Update node/store_test.go tests
- Add new fields to configs/edge.yaml
- Extend packages/go/config with new configuration options
- Update protobuf definitions in runtime.proto and generated code
This commit is contained in:
toki 2026-06-15 11:01:21 +09:00
parent 28945ee742
commit 76a8ee5b3b
13 changed files with 761 additions and 111 deletions

View file

@ -42,43 +42,44 @@ task=m-lemonade-provider-serving-validation/01_config_contract, plan=0, tag=API
| 항목 | 완료 여부 |
|------|---------|
| [API-1] OpenAI-compatible config contract | [ ] |
| [API-2] Typed proto payload and mapper | [ ] |
| [API-3] Edge validation and example config | [ ] |
| [API-1] OpenAI-compatible config contract | [x] |
| [API-2] Typed proto payload and mapper | [x] |
| [API-3] Edge validation and example config | [x] |
## 구현 체크리스트
- [ ] `packages/go/config`에 OpenAI-compatible inference server 설정 계약을 추가하고 normalization/validation 테스트를 작성한다.
- [ ] `proto/iop/runtime.proto`에 typed `OpenAICompatAdapterConfig`를 추가하고 `make proto`로 Go 생성물을 갱신한다.
- [ ] Edge config payload mapper와 config validation, `configs/edge.yaml` 예시를 새 계약에 맞춘다.
- [ ] 계획의 중간 검증과 최종 검증 명령을 실행하고 실제 출력을 review stub에 붙인다.
- [ ] CODE_REVIEW-*-G??.md의 구현 에이전트 소유 섹션을 실제 구현 내용과 검증 출력으로 채운다. 이 항목이 완료되기 전에는 구현이 완료된 것이 아니다.
- [x] `packages/go/config`에 OpenAI-compatible inference server 설정 계약을 추가하고 normalization/validation 테스트를 작성한다.
- [x] `proto/iop/runtime.proto`에 typed `OpenAICompatAdapterConfig`를 추가하고 `make proto`로 Go 생성물을 갱신한다.
- [x] Edge config payload mapper와 config validation, `configs/edge.yaml` 예시를 새 계약에 맞춘다.
- [x] 계획의 중간 검증과 최종 검증 명령을 실행하고 실제 출력을 review stub에 붙인다.
- [x] CODE_REVIEW-*-G??.md의 구현 에이전트 소유 섹션을 실제 구현 내용과 검증 출력으로 채운다. 이 항목이 완료되기 전에는 구현이 완료된 것이 아니다.
## 코드리뷰 전용 체크리스트
> **[REVIEW AGENT ONLY]** 이 체크리스트는 코드리뷰 에이전트만 사용한다.
> 구현 에이전트는 이 섹션을 수정하거나 체크하지 않는다.
- [ ] `코드리뷰 결과`에 `PASS`, `WARN`, `FAIL` 중 하나의 판정을 append한다.
- [ ] 판정과 `차원별 평가`, Required/Suggested/Nit 분류가 서로 일치한다.
- [ ] active `CODE_REVIEW-*-G??.md`를 `code_review_cloud_G05_N.log`로 아카이브한다.
- [ ] active `PLAN-*-G??.md`를 `plan_cloud_G05_M.log`로 아카이브한다.
- [ ] `.gitignore`의 Agent-Ops 관리 block이 `agent-task/**/*.md`와 `agent-task/**/*.log`를 unignore하고 `agent-roadmap/current.md`를 ignore하는지 확인한다.
- [ ] PASS이면 `agent-ops/skills/common/code-review/templates/complete-log-template.md` 기준으로 `complete.log`를 작성하고 active `.md` 파일을 남기지 않는다.
- [ ] PASS이면 active task 디렉터리 `agent-task/m-lemonade-provider-serving-validation/01_config_contract/`를 `agent-task/archive/YYYY/MM/m-lemonade-provider-serving-validation/01_config_contract/`로 이동하고 최종 archive 경로에서 이 체크리스트를 갱신한다.
- [ ] PASS이고 task group이 `m-<milestone-slug>`이면 런타임이 읽을 완료 이벤트 메타데이터를 보고하고, roadmap 수정이나 `update-roadmap` 직접 호출을 하지 않는다.
- [ ] PASS split 작업이면 이동 후 빈 active parent `agent-task/m-lemonade-provider-serving-validation/`를 제거하거나, 남은 sibling/file이 있어 유지했다고 확인한다.
- [x] `코드리뷰 결과`에 `PASS`, `WARN`, `FAIL` 중 하나의 판정을 append한다.
- [x] 판정과 `차원별 평가`, Required/Suggested/Nit 분류가 서로 일치한다.
- [x] active `CODE_REVIEW-*-G??.md`를 `code_review_cloud_G05_N.log`로 아카이브한다.
- [x] active `PLAN-*-G??.md`를 `plan_cloud_G05_M.log`로 아카이브한다.
- [x] `.gitignore`의 Agent-Ops 관리 block이 `agent-task/**/*.md`와 `agent-task/**/*.log`를 unignore하고 `agent-roadmap/current.md`를 ignore하는지 확인한다.
- [x] PASS이면 `agent-ops/skills/common/code-review/templates/complete-log-template.md` 기준으로 `complete.log`를 작성하고 active `.md` 파일을 남기지 않는다.
- [x] PASS이면 active task 디렉터리 `agent-task/m-lemonade-provider-serving-validation/01_config_contract/`를 `agent-task/archive/YYYY/MM/m-lemonade-provider-serving-validation/01_config_contract/`로 이동하고 최종 archive 경로에서 이 체크리스트를 갱신한다.
- [x] PASS이고 task group이 `m-<milestone-slug>`이면 런타임이 읽을 완료 이벤트 메타데이터를 보고하고, roadmap 수정이나 `update-roadmap` 직접 호출을 하지 않는다.
- [x] PASS split 작업이면 이동 후 빈 active parent `agent-task/m-lemonade-provider-serving-validation/`를 제거하거나, 남은 sibling/file이 있어 유지했다고 확인한다.
- [ ] WARN/FAIL이고 user-review gate가 트리거되지 않았으면 다음 active `PLAN-cloud-G05.md`와 `CODE_REVIEW-cloud-G05.md`를 작성하고 `complete.log`를 작성하지 않는다.
- [ ] USER_REVIEW이면 `agent-ops/skills/common/code-review/templates/user-review-template.md` 기준으로 `USER_REVIEW.md`를 작성하고 active `PLAN-*.md`, `CODE_REVIEW-*.md`, `complete.log`를 남기지 않는다.
- [ ] USER_REVIEW가 사용자 결정으로 완료/PASS 해소되면 `USER_REVIEW.md`를 해소 상태로 갱신하고 `complete.log`를 작성한 뒤 task directory를 archive로 이동한다.
## 계획 대비 변경 사항
_구현 에이전트가 계획과 다르게 구현한 부분을 이유와 함께 기록한다._
TestLoadEdge_OpenAICompatInstances 테스트 수행 시, viper의 디코더 특성에 의해 yaml로 지정된 `headers`의 key가 소문자로 normalization(e.g., `authorization`)되는 동작을 확인하여, 테스트 코드 내 검증 방식을 소문자 key 검증으로 수정하였습니다.
## 주요 설계 결정
_구현 에이전트가 주요 설계 결정 사항을 기록한다._
1. `openai_compat` adapter에 대한 configuration 필드들이 downstream으로 안전하게 parsing/unmarshaling 될 수 있도록 `OpenAICompatConf`와 `OpenAICompatInstanceConf` 구조체를 정의하고, `normalizeAdapters` 함수 내에서 unique name 검사 및 queue validation logic을 적용하였습니다.
2. `runtime.proto` 에 `OpenAICompatAdapterConfig` 메시지를 추가하고 oneof 필드(`openai_compat`)에 index `10`으로 등록하여, 기존 tag와의 충돌을 피하고 serialized payload가 edge에서 node로 원활히 mapper를 통해 전달되도록 하였습니다.
## 사용자 리뷰 요청
@ -101,43 +102,52 @@ _기본값은 `없음`이다. 구현 중 사용자 결정, 사용자 소유 외
## 검증 결과
_구현 에이전트가 각 중간 검증 및 최종 검증 명령 실행 후 출력을 여기에 붙여 넣는다._
필수 규칙:
- 검증 명령은 고정된 계약이다. 임의로 대체하지 않는다.
- 대체가 필요하면 `계획 대비 변경 사항`에 이유와 대체 명령을 기록한다.
- `검증 결과`에는 실제 stdout/stderr를 붙여 넣는다.
- 사용자 리뷰 요청으로 명령을 끝까지 실행하지 못했다면 `사용자 리뷰 요청`에 실행한 명령, 실제 출력, 미실행 명령의 사유를 기록한다.
- mobile/UI hang, timeout, 또는 2분 무진행은 blind retry를 중단하고 focused rerun 명령과 screenshot/window/UI-tree evidence path를 남기며, 불가능하면 정확한 사유를 남긴다.
### API-1 중간 검증
```bash
$ go test -count=1 ./packages/go/config
(output)
ok iop/packages/go/config 0.012s
```
### API-2 중간 검증
```bash
$ make proto
(output)
protoc \
--go_out=. \
--go_opt=module=iop \
--proto_path=. \
proto/iop/runtime.proto \
proto/iop/node.proto \
proto/iop/control.proto \
proto/iop/job.proto
$ go test -count=1 ./apps/edge/internal/node ./proto/gen/...
(output)
ok iop/apps/edge/internal/node 0.004s
? iop/proto/gen/iop [no test files]
```
### API-3 중간 검증
```bash
$ go test -count=1 ./apps/edge/internal/edgecmd
(output)
ok iop/apps/edge/internal/edgecmd 0.015s
```
### 최종 검증
```bash
$ make proto
(output)
protoc \
--go_out=. \
--go_opt=module=iop \
--proto_path=. \
proto/iop/runtime.proto \
proto/iop/node.proto \
proto/iop/control.proto \
proto/iop/job.proto
$ go test -count=1 ./packages/go/config ./apps/edge/internal/node ./apps/edge/internal/edgecmd ./proto/gen/...
(output)
ok iop/packages/go/config 0.014s
ok iop/apps/edge/internal/node 0.004s
ok iop/apps/edge/internal/edgecmd 0.006s
? iop/proto/gen/iop [no test files]
$ git diff --check
(output)
(출력 없음 - 성공)
```
---
@ -145,3 +155,24 @@ $ git diff --check
> **[IMPLEMENTING AGENT — BEFORE SAVING] Have you filled in every implementation-owned section: completion table, implementation checklist, changes from plan, design decisions, and verification output?**
> If anything is blank, go back and fill it in before saving this file.
> Leave review-agent-only sections unchanged.
## 코드리뷰 결과
- 리뷰 일시: 2026-06-15
- 종합 판정: PASS
- 차원별 평가:
- Correctness: Pass
- Completeness: Pass
- Test coverage: Pass
- API contract: Pass
- Code quality: Pass
- Plan deviation: Pass
- Verification trust: Pass
- 발견된 문제: 없음
- 리뷰 중 직접 수정:
- `packages/go/config/config.go`, `apps/edge/internal/node/mapper.go`의 normalization 관련 주석을 새 `OpenAICompatInstances` 계약과 맞췄다.
- 계획의 요구 테스트명과 일치하도록 `TestBuildConfigPayload_MultiOpenAICompatInstances`를 `TestBuildConfigPayload_OpenAICompatInstances`로 정리했다.
- 검증:
- `make proto && go test -count=1 ./packages/go/config ./apps/edge/internal/node ./apps/edge/internal/edgecmd ./proto/gen/... && git diff --check`
- `go test -count=1 ./apps/node/internal/adapters`
- 다음 단계: PASS이므로 `complete.log`를 작성하고 task directory를 archive로 이동한다.

View file

@ -0,0 +1,43 @@
# Complete - m-lemonade-provider-serving-validation/01_config_contract
## 완료 일시
2026-06-15
## 요약
Lemonade/OpenAI-compatible config and proto payload contract review completed in 1 loop with final verdict PASS.
## 루프 이력
| Plan | Review | Verdict | 메모 |
|------|--------|---------|------|
| `agent-task/archive/2026/06/m-lemonade-provider-serving-validation/01_config_contract/plan_cloud_G05_0.log` | `agent-task/archive/2026/06/m-lemonade-provider-serving-validation/01_config_contract/code_review_cloud_G05_0.log` | PASS | OpenAI-compatible config structs, proto oneof, Edge mapper/validation, and examples matched the plan; review repaired only non-behavioral comment/test-name drift. |
## 구현/정리 내용
- Added `openai_compat` and `openai_compat_instances` config contracts with normalization, duplicate-name checks, and queue validation.
- Added typed `OpenAICompatAdapterConfig` protobuf payload and Edge `BuildConfigPayload` mapping.
- Added Edge config validation and commented Lemonade/OpenAI-compatible examples in `configs/edge.yaml`.
- Added focused tests for config loading/normalization, payload mapping, store normalization, and Edge config validation.
- Review-time cleanup aligned normalization comments and the OpenAI-compatible mapper test name with the plan.
## 최종 검증
- `make proto && go test -count=1 ./packages/go/config ./apps/edge/internal/node ./apps/edge/internal/edgecmd ./proto/gen/... && git diff --check` - PASS; proto regenerated, changed packages passed, generated proto package had no test files, and diff whitespace check succeeded.
- `go test -count=1 ./apps/node/internal/adapters` - PASS; existing Node adapter payload consumers still compiled and passed.
## Roadmap Completion
- Milestone: `agent-roadmap/phase/inference-provider-extension/milestones/lemonade-provider-serving-validation.md`
- Completed task ids:
- `config-contract`: PASS; evidence=`agent-task/archive/2026/06/m-lemonade-provider-serving-validation/01_config_contract/plan_cloud_G05_0.log`, `agent-task/archive/2026/06/m-lemonade-provider-serving-validation/01_config_contract/code_review_cloud_G05_0.log`; verification=`make proto && go test -count=1 ./packages/go/config ./apps/edge/internal/node ./apps/edge/internal/edgecmd ./proto/gen/... && git diff --check`
- Not completed task ids: 없음
## 잔여 Nit
- 없음
## 후속 작업
- 없음

View file

@ -149,6 +149,18 @@ func anyVllmEnabled(a config.AdaptersConf) bool {
return false
}
func anyOpenAICompatEnabled(a config.AdaptersConf) bool {
if a.OpenAICompat.Enabled {
return true
}
for _, inst := range a.OpenAICompatInstances {
if inst.Enabled {
return true
}
}
return false
}
func validateEdgeConfig(cfg *config.EdgeConfig) error {
_, err := node.LoadFromConfig(cfg.Nodes)
if err != nil {
@ -166,7 +178,8 @@ func validateEdgeConfig(cfg *config.EdgeConfig) error {
hasOllama := anyOllamaEnabled(n.Adapters)
hasVllm := anyVllmEnabled(n.Adapters)
if !hasOllama && !n.Adapters.CLI.Enabled && !hasVllm {
hasOpenAICompat := anyOpenAICompatEnabled(n.Adapters)
if !hasOllama && !n.Adapters.CLI.Enabled && !hasVllm && !hasOpenAICompat {
return fmt.Errorf("node %q: at least one adapter must be enabled", name)
}
if n.Adapters.Ollama.Enabled && n.Adapters.Ollama.BaseURL == "" {
@ -185,6 +198,14 @@ func validateEdgeConfig(cfg *config.EdgeConfig) error {
return fmt.Errorf("node %q: vllm_instances[%d] %q: endpoint must not be empty", name, j, inst.Name)
}
}
if n.Adapters.OpenAICompat.Enabled && n.Adapters.OpenAICompat.Endpoint == "" {
return fmt.Errorf("node %q: openai_compat adapter endpoint must not be empty", name)
}
for j, inst := range n.Adapters.OpenAICompatInstances {
if inst.Enabled && inst.Endpoint == "" {
return fmt.Errorf("node %q: openai_compat_instances[%d] %q: endpoint must not be empty", name, j, inst.Name)
}
}
if n.Runtime.Concurrency < 0 {
return fmt.Errorf("node %q: runtime concurrency must be non-negative", name)
}

View file

@ -303,3 +303,63 @@ func TestValidateEdgeConfig_OllamaInstanceEnabledAccepted(t *testing.T) {
t.Fatalf("unexpected error: %v", err)
}
}
func TestValidateEdgeConfig_OpenAICompatInstanceEnabled(t *testing.T) {
cfg := &config.EdgeConfig{
Nodes: []config.NodeDefinition{
{
ID: "node-1",
Alias: "test",
Token: "tok",
Adapters: config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "lemonade", Enabled: true, Endpoint: "http://127.0.0.1:13305"},
},
},
},
},
}
if err := validateEdgeConfig(cfg); err != nil {
t.Fatalf("unexpected error: %v", err)
}
}
func TestValidateEdgeConfig_OpenAICompatInstanceEmptyEndpointRejected(t *testing.T) {
cfg := &config.EdgeConfig{
Nodes: []config.NodeDefinition{
{
ID: "node-1",
Alias: "test",
Token: "tok",
Adapters: config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "lemonade", Enabled: true, Endpoint: ""},
},
},
},
},
}
err := validateEdgeConfig(cfg)
if err == nil {
t.Fatal("expected error for enabled openai_compat instance with empty endpoint")
}
}
func TestValidateEdgeConfig_OpenAICompatLegacyEmptyEndpointRejected(t *testing.T) {
cfg := &config.EdgeConfig{
Nodes: []config.NodeDefinition{
{
ID: "node-1",
Alias: "test",
Token: "tok",
Adapters: config.AdaptersConf{
OpenAICompat: config.OpenAICompatConf{Enabled: true, Endpoint: ""},
},
},
},
}
err := validateEdgeConfig(cfg)
if err == nil {
t.Fatal("expected error for enabled legacy openai_compat with empty endpoint")
}
}

View file

@ -10,9 +10,10 @@ import (
// BuildConfigPayload converts a NodeRecord's adapter config to the proto payload
// sent to node during registration.
//
// After config normalisation, OllamaInstances and VllmInstances already contain
// the promoted legacy single-instance entries, so only the slice fields are
// iterated here. Duplicate instance names produce an error.
// After config normalisation, OllamaInstances, VllmInstances, and
// OpenAICompatInstances already contain the promoted legacy single-instance
// entries, so only the slice fields are iterated here. Duplicate instance names
// produce an error.
func BuildConfigPayload(rec *NodeRecord) (*iop.NodeConfigPayload, error) {
payload := &iop.NodeConfigPayload{
Runtime: &iop.NodeRuntimeConfig{
@ -80,6 +81,33 @@ func BuildConfigPayload(rec *NodeRecord) (*iop.NodeConfigPayload, error) {
})
}
for _, inst := range rec.Adapters.OpenAICompatInstances {
if !inst.Enabled {
continue
}
key := "openai_compat:" + inst.Name
if _, dup := seen[key]; dup {
return nil, fmt.Errorf("duplicate openai_compat instance name %q", inst.Name)
}
seen[key] = struct{}{}
payload.Adapters = append(payload.Adapters, &iop.AdapterConfig{
Type: "openai_compat",
Enabled: true,
Name: inst.Name,
Config: &iop.AdapterConfig_OpenaiCompat{
OpenaiCompat: &iop.OpenAICompatAdapterConfig{
Provider: inst.Provider,
Endpoint: inst.Endpoint,
Headers: inst.Headers,
Capacity: int32(inst.Capacity),
MaxQueue: int32(inst.MaxQueue),
QueueTimeoutMs: int32(inst.QueueTimeoutMS),
RequestTimeoutMs: int32(inst.RequestTimeoutMS),
},
},
})
}
if rec.Adapters.CLI.Enabled {
profiles := make(map[string]*iop.CLIProfileConfig, len(rec.Adapters.CLI.Profiles))
for name, p := range rec.Adapters.CLI.Profiles {

View file

@ -441,3 +441,86 @@ func TestBuildConfigPayload_MockUsesTypedConfig(t *testing.T) {
t.Fatal("mock adapter must use typed MockAdapterConfig oneof")
}
}
func TestBuildConfigPayload_OpenAICompatInstances(t *testing.T) {
rec := &edgenode.NodeRecord{
ID: "node-1",
Alias: "test",
Token: "token",
Adapters: config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "lemonade", Enabled: true, Provider: "lemonade", Endpoint: "http://127.0.0.1:13305", Headers: map[string]string{"authorization": "Bearer test"}, Capacity: 4, MaxQueue: 10, QueueTimeoutMS: 1500, RequestTimeoutMS: 30000},
{Name: "openai-api", Enabled: true, Provider: "openai", Endpoint: "https://api.openai.com/v1", Headers: map[string]string{"authorization": "Bearer sk-test"}, Capacity: 8, MaxQueue: 20, QueueTimeoutMS: 2500, RequestTimeoutMS: 60000},
},
},
Runtime: config.RuntimeConf{WorkspaceRoot: "/tmp/ws"},
}
payload, err := edgenode.BuildConfigPayload(rec)
if err != nil {
t.Fatalf("BuildConfigPayload failed: %v", err)
}
var oaiAdapters []*iop.AdapterConfig
for _, a := range payload.Adapters {
if a.Type == "openai_compat" {
oaiAdapters = append(oaiAdapters, a)
}
}
if len(oaiAdapters) != 2 {
t.Fatalf("expected 2 openai_compat adapters, got %d", len(oaiAdapters))
}
names := map[string]bool{}
for _, a := range oaiAdapters {
names[a.Name] = true
if a.GetOpenaiCompat() == nil {
t.Errorf("instance %q: expected openai_compat typed config", a.Name)
}
switch a.Name {
case "lemonade":
oai := a.GetOpenaiCompat()
if oai.GetProvider() != "lemonade" || oai.GetEndpoint() != "http://127.0.0.1:13305" {
t.Errorf("lemonade provider/endpoint mismatch: %+v", oai)
}
if oai.GetHeaders()["authorization"] != "Bearer test" {
t.Errorf("lemonade headers mismatch: %+v", oai.GetHeaders())
}
if oai.GetCapacity() != 4 || oai.GetMaxQueue() != 10 || oai.GetQueueTimeoutMs() != 1500 || oai.GetRequestTimeoutMs() != 30000 {
t.Errorf("lemonade queue config mismatch: %+v", oai)
}
case "openai-api":
oai := a.GetOpenaiCompat()
if oai.GetProvider() != "openai" || oai.GetEndpoint() != "https://api.openai.com/v1" {
t.Errorf("openai-api provider/endpoint mismatch: %+v", oai)
}
if oai.GetHeaders()["authorization"] != "Bearer sk-test" {
t.Errorf("openai-api headers mismatch: %+v", oai.GetHeaders())
}
if oai.GetCapacity() != 8 || oai.GetMaxQueue() != 20 || oai.GetQueueTimeoutMs() != 2500 || oai.GetRequestTimeoutMs() != 60000 {
t.Errorf("openai-api queue config mismatch: %+v", oai)
}
}
}
if !names["lemonade"] || !names["openai-api"] {
t.Errorf("expected instance names lemonade and openai-api, got %v", names)
}
}
func TestBuildConfigPayload_DuplicateOpenAICompatNameError(t *testing.T) {
rec := &edgenode.NodeRecord{
ID: "node-1",
Alias: "test",
Token: "token",
Adapters: config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "dup", Enabled: true, Endpoint: "http://127.0.0.1:13305"},
{Name: "dup", Enabled: true, Endpoint: "http://127.0.0.2:13305"},
},
},
}
_, err := edgenode.BuildConfigPayload(rec)
if err == nil {
t.Fatal("expected error for duplicate openai_compat instance name")
}
}

View file

@ -230,3 +230,44 @@ func TestLoadFromConfig_AutoIDUnique(t *testing.T) {
t.Fatalf("expected unique IDs, both are %q", recA.ID)
}
}
func TestLoadFromConfig_LegacyOpenAICompatInPayload(t *testing.T) {
store, err := edgenode.LoadFromConfig([]config.NodeDefinition{
{
ID: "node-oai",
Alias: "oai-node",
Token: "token-oai",
Adapters: config.AdaptersConf{
OpenAICompat: config.OpenAICompatConf{
Enabled: true,
Provider: "lemonade",
Endpoint: "http://127.0.0.1:13305",
Headers: map[string]string{"authorization": "Bearer token"},
},
},
},
})
if err != nil {
t.Fatalf("load from config: %v", err)
}
rec, ok := store.FindByToken("token-oai")
if !ok {
t.Fatal("expected record for token-oai")
}
payload, err := edgenode.BuildConfigPayload(rec)
if err != nil {
t.Fatalf("build config payload: %v", err)
}
var found bool
for _, a := range payload.GetAdapters() {
if a.GetType() == "openai_compat" {
oai := a.GetOpenaiCompat()
if oai != nil && oai.GetProvider() == "lemonade" && oai.GetEndpoint() == "http://127.0.0.1:13305" && oai.GetHeaders()["authorization"] == "Bearer token" {
found = true
}
}
}
if !found {
t.Fatalf("expected openai_compat adapter in payload with correct config, got %v", payload.GetAdapters())
}
}

View file

@ -69,6 +69,10 @@ openai:
# adapter: "vllm"
# target: "qwen3-72b"
# node: "node-gpu-01"
# - model: "lemonade"
# adapter: "openai_compat"
# target: "lemonade-served-model"
# node: "node-lemonade-01"
session_id: "openai"
timeout_sec: 120
strict_output: true
@ -184,6 +188,17 @@ nodes:
# claude:
# command: "claude"
# mode: "antigravity-print"
# openai_compat_instances:
# - name: "lemonade"
# enabled: true
# provider: "lemonade"
# endpoint: "http://127.0.0.1:13305"
# headers:
# Authorization: "Bearer <lemonade-api-key-placeholder>"
# capacity: 4
# max_queue: 16
# queue_timeout_ms: 30000
# request_timeout_ms: 300000
# runtime:
# concurrency: 4
# workspace_root: "/workspace"

View file

@ -181,17 +181,19 @@ type MetricsConf struct {
type AdaptersConf struct {
// Legacy single-instance fields. When non-empty they are normalised into
// OllamaInstances / VllmInstances during config load so that all downstream
// code only needs to inspect the slice fields.
Ollama OllamaConf `mapstructure:"ollama" yaml:"ollama"`
Vllm VllmConf `mapstructure:"vllm" yaml:"vllm"`
CLI CLIConf `mapstructure:"cli" yaml:"cli"`
// OllamaInstances / VllmInstances / OpenAICompatInstances during config load
// so that all downstream code only needs to inspect the slice fields.
Ollama OllamaConf `mapstructure:"ollama" yaml:"ollama"`
Vllm VllmConf `mapstructure:"vllm" yaml:"vllm"`
OpenAICompat OpenAICompatConf `mapstructure:"openai_compat" yaml:"openai_compat"`
CLI CLIConf `mapstructure:"cli" yaml:"cli"`
// Multi-instance collections. Each entry carries a unique Name that acts as
// the stable adapter instance identity within the node. Names must be unique
// within each type collection; duplicate names are rejected at load time.
OllamaInstances []OllamaInstanceConf `mapstructure:"ollama_instances" yaml:"ollama_instances,omitempty"`
VllmInstances []VllmInstanceConf `mapstructure:"vllm_instances" yaml:"vllm_instances,omitempty"`
OllamaInstances []OllamaInstanceConf `mapstructure:"ollama_instances" yaml:"ollama_instances,omitempty"`
VllmInstances []VllmInstanceConf `mapstructure:"vllm_instances" yaml:"vllm_instances,omitempty"`
OpenAICompatInstances []OpenAICompatInstanceConf `mapstructure:"openai_compat_instances" yaml:"openai_compat_instances,omitempty"`
}
// OllamaInstanceConf is one named Ollama engine instance within a node.
@ -217,6 +219,29 @@ type VllmInstanceConf struct {
RequestTimeoutMS int `mapstructure:"request_timeout_ms" yaml:"request_timeout_ms"`
}
type OpenAICompatConf struct {
Enabled bool `mapstructure:"enabled" yaml:"enabled"`
Provider string `mapstructure:"provider" yaml:"provider"`
Endpoint string `mapstructure:"endpoint" yaml:"endpoint"`
Headers map[string]string `mapstructure:"headers" yaml:"headers"`
Capacity int `mapstructure:"capacity" yaml:"capacity"`
MaxQueue int `mapstructure:"max_queue" yaml:"max_queue"`
QueueTimeoutMS int `mapstructure:"queue_timeout_ms" yaml:"queue_timeout_ms"`
RequestTimeoutMS int `mapstructure:"request_timeout_ms" yaml:"request_timeout_ms"`
}
type OpenAICompatInstanceConf struct {
Name string `mapstructure:"name" yaml:"name"`
Enabled bool `mapstructure:"enabled" yaml:"enabled"`
Provider string `mapstructure:"provider" yaml:"provider"`
Endpoint string `mapstructure:"endpoint" yaml:"endpoint"`
Headers map[string]string `mapstructure:"headers" yaml:"headers"`
Capacity int `mapstructure:"capacity" yaml:"capacity"`
MaxQueue int `mapstructure:"max_queue" yaml:"max_queue"`
QueueTimeoutMS int `mapstructure:"queue_timeout_ms" yaml:"queue_timeout_ms"`
RequestTimeoutMS int `mapstructure:"request_timeout_ms" yaml:"request_timeout_ms"`
}
type OllamaConf struct {
Enabled bool `mapstructure:"enabled" yaml:"enabled"`
BaseURL string `mapstructure:"base_url" yaml:"base_url"`
@ -357,6 +382,24 @@ func normalizeAdapters(a *AdaptersConf) error {
return fmt.Errorf("vllm: legacy field conflicts with explicit instance %q: enabled/endpoint/capacity/max_queue/queue_timeout_ms/request_timeout_ms mismatch", "vllm")
}
}
if a.OpenAICompat.Enabled {
existing := findOpenAICompatInstance(a.OpenAICompatInstances, "openai_compat")
if existing == nil {
a.OpenAICompatInstances = append([]OpenAICompatInstanceConf{{
Name: "openai_compat",
Enabled: a.OpenAICompat.Enabled,
Provider: a.OpenAICompat.Provider,
Endpoint: a.OpenAICompat.Endpoint,
Headers: a.OpenAICompat.Headers,
Capacity: a.OpenAICompat.Capacity,
MaxQueue: a.OpenAICompat.MaxQueue,
QueueTimeoutMS: a.OpenAICompat.QueueTimeoutMS,
RequestTimeoutMS: a.OpenAICompat.RequestTimeoutMS,
}}, a.OpenAICompatInstances...)
} else if !sameOpenAICompatInstance(*existing, a.OpenAICompat) {
return fmt.Errorf("openai_compat: legacy field conflicts with explicit instance %q: enabled/provider/endpoint/headers/capacity/max_queue/queue_timeout_ms/request_timeout_ms mismatch", "openai_compat")
}
}
if err := checkUniqueNames("ollama_instances", func(i int) string { return a.OllamaInstances[i].Name }, len(a.OllamaInstances)); err != nil {
return err
@ -364,6 +407,9 @@ func normalizeAdapters(a *AdaptersConf) error {
if err := checkUniqueNames("vllm_instances", func(i int) string { return a.VllmInstances[i].Name }, len(a.VllmInstances)); err != nil {
return err
}
if err := checkUniqueNames("openai_compat_instances", func(i int) string { return a.OpenAICompatInstances[i].Name }, len(a.OpenAICompatInstances)); err != nil {
return err
}
for i, inst := range a.OllamaInstances {
field := fmt.Sprintf("ollama_instances[%d]", i)
if err := validateProviderQueueConfig(field, inst.Capacity, inst.MaxQueue, inst.QueueTimeoutMS, inst.RequestTimeoutMS); err != nil {
@ -376,6 +422,12 @@ func normalizeAdapters(a *AdaptersConf) error {
return err
}
}
for i, inst := range a.OpenAICompatInstances {
field := fmt.Sprintf("openai_compat_instances[%d]", i)
if err := validateProviderQueueConfig(field, inst.Capacity, inst.MaxQueue, inst.QueueTimeoutMS, inst.RequestTimeoutMS); err != nil {
return err
}
}
return nil
}
@ -432,6 +484,36 @@ func findVllmInstance(instances []VllmInstanceConf, name string) *VllmInstanceCo
return nil
}
func sameOpenAICompatInstance(inst OpenAICompatInstanceConf, legacy OpenAICompatConf) bool {
if inst.Enabled != legacy.Enabled ||
inst.Provider != legacy.Provider ||
inst.Endpoint != legacy.Endpoint ||
inst.Capacity != legacy.Capacity ||
inst.MaxQueue != legacy.MaxQueue ||
inst.QueueTimeoutMS != legacy.QueueTimeoutMS ||
inst.RequestTimeoutMS != legacy.RequestTimeoutMS {
return false
}
if len(inst.Headers) != len(legacy.Headers) {
return false
}
for k, v := range inst.Headers {
if lv, ok := legacy.Headers[k]; !ok || lv != v {
return false
}
}
return true
}
func findOpenAICompatInstance(instances []OpenAICompatInstanceConf, name string) *OpenAICompatInstanceConf {
for i := range instances {
if instances[i].Name == name {
return &instances[i]
}
}
return nil
}
// validateOpenAIRoutes rejects duplicate and empty model ids in the route catalog.
func validateOpenAIRoutes(routes []OpenAIRouteEntry) error {
seen := make(map[string]struct{}, len(routes))

View file

@ -1461,3 +1461,112 @@ control_plane:
t.Fatalf("expected control_plane.reconnect_interval_sec=10, got %d", cfg.ControlPlane.ReconnectIntervalSec)
}
}
func TestLoadEdge_OpenAICompatInstances(t *testing.T) {
dir := t.TempDir()
f := filepath.Join(dir, "edge.yaml")
yaml := `
server:
listen: "0.0.0.0:9090"
nodes:
- alias: "lemonade-node"
adapters:
openai_compat_instances:
- name: "lemonade"
enabled: true
provider: "lemonade"
endpoint: "http://127.0.0.1:13305"
headers:
Authorization: "Bearer test-key"
capacity: 4
max_queue: 10
queue_timeout_ms: 1500
request_timeout_ms: 30000
`
if err := os.WriteFile(f, []byte(yaml), 0o600); err != nil {
t.Fatalf("write yaml: %v", err)
}
cfg, err := config.LoadEdge(f)
if err != nil {
t.Fatalf("load: %v", err)
}
if len(cfg.Nodes) == 0 {
t.Fatal("expected 1 node")
}
insts := cfg.Nodes[0].Adapters.OpenAICompatInstances
if len(insts) != 1 {
t.Fatalf("expected 1 openai_compat instance, got %d", len(insts))
}
inst := insts[0]
if inst.Name != "lemonade" || inst.Provider != "lemonade" || inst.Endpoint != "http://127.0.0.1:13305" {
t.Errorf("unexpected instance config: %+v", inst)
}
if inst.Headers["authorization"] != "Bearer test-key" {
t.Errorf("expected header authorization Bearer test-key, got %q", inst.Headers["authorization"])
}
if inst.Capacity != 4 || inst.MaxQueue != 10 || inst.QueueTimeoutMS != 1500 || inst.RequestTimeoutMS != 30000 {
t.Errorf("unexpected queue config: %+v", inst)
}
}
func TestNormalizeAdapters_OpenAICompatIdempotent(t *testing.T) {
a := config.AdaptersConf{
OpenAICompat: config.OpenAICompatConf{
Enabled: true,
Provider: "lemonade",
Endpoint: "http://127.0.0.1:13305",
Headers: map[string]string{"X-Test": "val"},
},
}
if err := config.NormalizeAdapters(&a); err != nil {
t.Fatalf("first normalize: %v", err)
}
if len(a.OpenAICompatInstances) != 1 {
t.Fatalf("expected 1 instance after first normalize, got %d", len(a.OpenAICompatInstances))
}
if err := config.NormalizeAdapters(&a); err != nil {
t.Fatalf("second normalize (idempotent): %v", err)
}
if len(a.OpenAICompatInstances) != 1 {
t.Fatalf("expected still 1 instance after second normalize, got %d", len(a.OpenAICompatInstances))
}
}
func TestNormalizeAdapters_OpenAICompatConflictErrors(t *testing.T) {
a := config.AdaptersConf{
OpenAICompat: config.OpenAICompatConf{Enabled: true, Endpoint: "http://127.0.0.1:13305"},
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "openai_compat", Enabled: true, Endpoint: "http://other-host:13305"},
},
}
if err := config.NormalizeAdapters(&a); err == nil {
t.Fatal("expected error for conflicting legacy openai_compat and explicit instance with same name")
}
}
func TestNormalizeAdapters_OpenAICompatDuplicateNameRejects(t *testing.T) {
a := config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "same", Enabled: true, Endpoint: "http://127.0.0.1:13305"},
{Name: "same", Enabled: true, Endpoint: "http://127.0.0.2:13305"},
},
}
if err := config.NormalizeAdapters(&a); err == nil {
t.Fatal("expected error for duplicate openai_compat instance name")
}
}
func TestNormalizeAdapters_OpenAICompatQueueValidation(t *testing.T) {
a := config.AdaptersConf{
OpenAICompatInstances: []config.OpenAICompatInstanceConf{
{Name: "test", Enabled: true, Endpoint: "http://127.0.0.1:13305", Capacity: -1},
},
}
err := config.NormalizeAdapters(&a)
if err == nil {
t.Fatal("expected negative provider queue config error")
}
if !strings.Contains(err.Error(), "capacity") {
t.Fatalf("expected error to mention capacity, got %v", err)
}
}

View file

@ -1309,7 +1309,7 @@ func (x *NodeConfigPayload) GetRuntime() *NodeRuntimeConfig {
// multiple instances of the same adapter type each must have a unique name.
type AdapterConfig struct {
state protoimpl.MessageState `protogen:"open.v1"`
Type string `protobuf:"bytes,1,opt,name=type,proto3" json:"type,omitempty"` // "mock" | "ollama" | "vllm" | "cli"
Type string `protobuf:"bytes,1,opt,name=type,proto3" json:"type,omitempty"` // "mock" | "ollama" | "vllm" | "cli" | "openai_compat"
Enabled bool `protobuf:"varint,2,opt,name=enabled,proto3" json:"enabled,omitempty"`
Settings *structpb.Struct `protobuf:"bytes,3,opt,name=settings,proto3" json:"settings,omitempty"` // legacy/compat path; new adapters use oneof
// Types that are valid to be assigned to Config:
@ -1318,6 +1318,7 @@ type AdapterConfig struct {
// *AdapterConfig_Ollama
// *AdapterConfig_Vllm
// *AdapterConfig_Mock
// *AdapterConfig_OpenaiCompat
Config isAdapterConfig_Config `protobuf_oneof:"config"`
Name string `protobuf:"bytes,8,opt,name=name,proto3" json:"name,omitempty"` // stable instance identity; empty = legacy single-instance
Target string `protobuf:"bytes,9,opt,name=target,proto3" json:"target,omitempty"` // optional default target/route hint for this instance
@ -1419,6 +1420,15 @@ func (x *AdapterConfig) GetMock() *MockAdapterConfig {
return nil
}
func (x *AdapterConfig) GetOpenaiCompat() *OpenAICompatAdapterConfig {
if x != nil {
if x, ok := x.Config.(*AdapterConfig_OpenaiCompat); ok {
return x.OpenaiCompat
}
}
return nil
}
func (x *AdapterConfig) GetName() string {
if x != nil {
return x.Name
@ -1453,6 +1463,10 @@ type AdapterConfig_Mock struct {
Mock *MockAdapterConfig `protobuf:"bytes,7,opt,name=mock,proto3,oneof"`
}
type AdapterConfig_OpenaiCompat struct {
OpenaiCompat *OpenAICompatAdapterConfig `protobuf:"bytes,10,opt,name=openai_compat,json=openaiCompat,proto3,oneof"`
}
func (*AdapterConfig_Cli) isAdapterConfig_Config() {}
func (*AdapterConfig_Ollama) isAdapterConfig_Config() {}
@ -1461,6 +1475,8 @@ func (*AdapterConfig_Vllm) isAdapterConfig_Config() {}
func (*AdapterConfig_Mock) isAdapterConfig_Config() {}
func (*AdapterConfig_OpenaiCompat) isAdapterConfig_Config() {}
type MockAdapterConfig struct {
state protoimpl.MessageState `protogen:"open.v1"`
unknownFields protoimpl.UnknownFields
@ -1877,6 +1893,98 @@ func (x *VllmAdapterConfig) GetRequestTimeoutMs() int32 {
return 0
}
type OpenAICompatAdapterConfig struct {
state protoimpl.MessageState `protogen:"open.v1"`
Provider string `protobuf:"bytes,1,opt,name=provider,proto3" json:"provider,omitempty"`
Endpoint string `protobuf:"bytes,2,opt,name=endpoint,proto3" json:"endpoint,omitempty"`
Headers map[string]string `protobuf:"bytes,3,rep,name=headers,proto3" json:"headers,omitempty" protobuf_key:"bytes,1,opt,name=key" protobuf_val:"bytes,2,opt,name=value"`
Capacity int32 `protobuf:"varint,4,opt,name=capacity,proto3" json:"capacity,omitempty"`
MaxQueue int32 `protobuf:"varint,5,opt,name=max_queue,json=maxQueue,proto3" json:"max_queue,omitempty"`
QueueTimeoutMs int32 `protobuf:"varint,6,opt,name=queue_timeout_ms,json=queueTimeoutMs,proto3" json:"queue_timeout_ms,omitempty"`
RequestTimeoutMs int32 `protobuf:"varint,7,opt,name=request_timeout_ms,json=requestTimeoutMs,proto3" json:"request_timeout_ms,omitempty"`
unknownFields protoimpl.UnknownFields
sizeCache protoimpl.SizeCache
}
func (x *OpenAICompatAdapterConfig) Reset() {
*x = OpenAICompatAdapterConfig{}
mi := &file_proto_iop_runtime_proto_msgTypes[21]
ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
ms.StoreMessageInfo(mi)
}
func (x *OpenAICompatAdapterConfig) String() string {
return protoimpl.X.MessageStringOf(x)
}
func (*OpenAICompatAdapterConfig) ProtoMessage() {}
func (x *OpenAICompatAdapterConfig) ProtoReflect() protoreflect.Message {
mi := &file_proto_iop_runtime_proto_msgTypes[21]
if x != nil {
ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
if ms.LoadMessageInfo() == nil {
ms.StoreMessageInfo(mi)
}
return ms
}
return mi.MessageOf(x)
}
// Deprecated: Use OpenAICompatAdapterConfig.ProtoReflect.Descriptor instead.
func (*OpenAICompatAdapterConfig) Descriptor() ([]byte, []int) {
return file_proto_iop_runtime_proto_rawDescGZIP(), []int{21}
}
func (x *OpenAICompatAdapterConfig) GetProvider() string {
if x != nil {
return x.Provider
}
return ""
}
func (x *OpenAICompatAdapterConfig) GetEndpoint() string {
if x != nil {
return x.Endpoint
}
return ""
}
func (x *OpenAICompatAdapterConfig) GetHeaders() map[string]string {
if x != nil {
return x.Headers
}
return nil
}
func (x *OpenAICompatAdapterConfig) GetCapacity() int32 {
if x != nil {
return x.Capacity
}
return 0
}
func (x *OpenAICompatAdapterConfig) GetMaxQueue() int32 {
if x != nil {
return x.MaxQueue
}
return 0
}
func (x *OpenAICompatAdapterConfig) GetQueueTimeoutMs() int32 {
if x != nil {
return x.QueueTimeoutMs
}
return 0
}
func (x *OpenAICompatAdapterConfig) GetRequestTimeoutMs() int32 {
if x != nil {
return x.RequestTimeoutMs
}
return 0
}
// NodeRuntimeConfig is the runtime tuning pushed to the node.
type NodeRuntimeConfig struct {
state protoimpl.MessageState `protogen:"open.v1"`
@ -1888,7 +1996,7 @@ type NodeRuntimeConfig struct {
func (x *NodeRuntimeConfig) Reset() {
*x = NodeRuntimeConfig{}
mi := &file_proto_iop_runtime_proto_msgTypes[21]
mi := &file_proto_iop_runtime_proto_msgTypes[22]
ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
ms.StoreMessageInfo(mi)
}
@ -1900,7 +2008,7 @@ func (x *NodeRuntimeConfig) String() string {
func (*NodeRuntimeConfig) ProtoMessage() {}
func (x *NodeRuntimeConfig) ProtoReflect() protoreflect.Message {
mi := &file_proto_iop_runtime_proto_msgTypes[21]
mi := &file_proto_iop_runtime_proto_msgTypes[22]
if x != nil {
ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
if ms.LoadMessageInfo() == nil {
@ -1913,7 +2021,7 @@ func (x *NodeRuntimeConfig) ProtoReflect() protoreflect.Message {
// Deprecated: Use NodeRuntimeConfig.ProtoReflect.Descriptor instead.
func (*NodeRuntimeConfig) Descriptor() ([]byte, []int) {
return file_proto_iop_runtime_proto_rawDescGZIP(), []int{21}
return file_proto_iop_runtime_proto_rawDescGZIP(), []int{22}
}
func (x *NodeRuntimeConfig) GetConcurrency() int32 {
@ -2062,7 +2170,7 @@ const file_proto_iop_runtime_proto_rawDesc = "" +
"\x06config\x18\x05 \x01(\v2\x16.iop.NodeConfigPayloadR\x06config\"u\n" +
"\x11NodeConfigPayload\x12.\n" +
"\badapters\x18\x01 \x03(\v2\x12.iop.AdapterConfigR\badapters\x120\n" +
"\aruntime\x18\x02 \x01(\v2\x16.iop.NodeRuntimeConfigR\aruntime\"\xe3\x02\n" +
"\aruntime\x18\x02 \x01(\v2\x16.iop.NodeRuntimeConfigR\aruntime\"\xaa\x03\n" +
"\rAdapterConfig\x12\x12\n" +
"\x04type\x18\x01 \x01(\tR\x04type\x12\x18\n" +
"\aenabled\x18\x02 \x01(\bR\aenabled\x123\n" +
@ -2070,7 +2178,9 @@ const file_proto_iop_runtime_proto_rawDesc = "" +
"\x03cli\x18\x04 \x01(\v2\x15.iop.CLIAdapterConfigH\x00R\x03cli\x122\n" +
"\x06ollama\x18\x05 \x01(\v2\x18.iop.OllamaAdapterConfigH\x00R\x06ollama\x12,\n" +
"\x04vllm\x18\x06 \x01(\v2\x16.iop.VllmAdapterConfigH\x00R\x04vllm\x12,\n" +
"\x04mock\x18\a \x01(\v2\x16.iop.MockAdapterConfigH\x00R\x04mock\x12\x12\n" +
"\x04mock\x18\a \x01(\v2\x16.iop.MockAdapterConfigH\x00R\x04mock\x12E\n" +
"\ropenai_compat\x18\n" +
" \x01(\v2\x1e.iop.OpenAICompatAdapterConfigH\x00R\fopenaiCompat\x12\x12\n" +
"\x04name\x18\b \x01(\tR\x04name\x12\x16\n" +
"\x06target\x18\t \x01(\tR\x06targetB\b\n" +
"\x06config\"\x13\n" +
@ -2111,7 +2221,18 @@ const file_proto_iop_runtime_proto_rawDesc = "" +
"\bcapacity\x18\x02 \x01(\x05R\bcapacity\x12\x1b\n" +
"\tmax_queue\x18\x03 \x01(\x05R\bmaxQueue\x12(\n" +
"\x10queue_timeout_ms\x18\x04 \x01(\x05R\x0equeueTimeoutMs\x12,\n" +
"\x12request_timeout_ms\x18\x05 \x01(\x05R\x10requestTimeoutMs\"\\\n" +
"\x12request_timeout_ms\x18\x05 \x01(\x05R\x10requestTimeoutMs\"\xe7\x02\n" +
"\x19OpenAICompatAdapterConfig\x12\x1a\n" +
"\bprovider\x18\x01 \x01(\tR\bprovider\x12\x1a\n" +
"\bendpoint\x18\x02 \x01(\tR\bendpoint\x12E\n" +
"\aheaders\x18\x03 \x03(\v2+.iop.OpenAICompatAdapterConfig.HeadersEntryR\aheaders\x12\x1a\n" +
"\bcapacity\x18\x04 \x01(\x05R\bcapacity\x12\x1b\n" +
"\tmax_queue\x18\x05 \x01(\x05R\bmaxQueue\x12(\n" +
"\x10queue_timeout_ms\x18\x06 \x01(\x05R\x0equeueTimeoutMs\x12,\n" +
"\x12request_timeout_ms\x18\a \x01(\x05R\x10requestTimeoutMs\x1a:\n" +
"\fHeadersEntry\x12\x10\n" +
"\x03key\x18\x01 \x01(\tR\x03key\x12\x14\n" +
"\x05value\x18\x02 \x01(\tR\x05value:\x028\x01\"\\\n" +
"\x11NodeRuntimeConfig\x12 \n" +
"\vconcurrency\x18\x01 \x01(\x05R\vconcurrency\x12%\n" +
"\x0eworkspace_root\x18\x02 \x01(\tR\rworkspaceRoot*\x81\x01\n" +
@ -2144,74 +2265,78 @@ func file_proto_iop_runtime_proto_rawDescGZIP() []byte {
}
var file_proto_iop_runtime_proto_enumTypes = make([]protoimpl.EnumInfo, 3)
var file_proto_iop_runtime_proto_msgTypes = make([]protoimpl.MessageInfo, 29)
var file_proto_iop_runtime_proto_msgTypes = make([]protoimpl.MessageInfo, 31)
var file_proto_iop_runtime_proto_goTypes = []any{
(RunSessionMode)(0), // 0: iop.RunSessionMode
(CancelAction)(0), // 1: iop.CancelAction
(NodeCommandType)(0), // 2: iop.NodeCommandType
(*RunRequest)(nil), // 3: iop.RunRequest
(*RunEvent)(nil), // 4: iop.RunEvent
(*EdgeNodeEvent)(nil), // 5: iop.EdgeNodeEvent
(*Usage)(nil), // 6: iop.Usage
(*Heartbeat)(nil), // 7: iop.Heartbeat
(*CancelRequest)(nil), // 8: iop.CancelRequest
(*NodeCommandRequest)(nil), // 9: iop.NodeCommandRequest
(*NodeCommandResponse)(nil), // 10: iop.NodeCommandResponse
(*ProviderSnapshot)(nil), // 11: iop.ProviderSnapshot
(*AgentUsageStatus)(nil), // 12: iop.AgentUsageStatus
(*Error)(nil), // 13: iop.Error
(*RegisterRequest)(nil), // 14: iop.RegisterRequest
(*RegisterResponse)(nil), // 15: iop.RegisterResponse
(*NodeConfigPayload)(nil), // 16: iop.NodeConfigPayload
(*AdapterConfig)(nil), // 17: iop.AdapterConfig
(*MockAdapterConfig)(nil), // 18: iop.MockAdapterConfig
(*CLIAdapterConfig)(nil), // 19: iop.CLIAdapterConfig
(*CLIProfileConfig)(nil), // 20: iop.CLIProfileConfig
(*CLICompletionMarker)(nil), // 21: iop.CLICompletionMarker
(*OllamaAdapterConfig)(nil), // 22: iop.OllamaAdapterConfig
(*VllmAdapterConfig)(nil), // 23: iop.VllmAdapterConfig
(*NodeRuntimeConfig)(nil), // 24: iop.NodeRuntimeConfig
nil, // 25: iop.RunRequest.MetadataEntry
nil, // 26: iop.RunEvent.MetadataEntry
nil, // 27: iop.EdgeNodeEvent.MetadataEntry
nil, // 28: iop.NodeCommandRequest.MetadataEntry
nil, // 29: iop.NodeCommandResponse.ResultEntry
nil, // 30: iop.AgentUsageStatus.MetadataEntry
nil, // 31: iop.CLIAdapterConfig.ProfilesEntry
(*structpb.Struct)(nil), // 32: google.protobuf.Struct
(RunSessionMode)(0), // 0: iop.RunSessionMode
(CancelAction)(0), // 1: iop.CancelAction
(NodeCommandType)(0), // 2: iop.NodeCommandType
(*RunRequest)(nil), // 3: iop.RunRequest
(*RunEvent)(nil), // 4: iop.RunEvent
(*EdgeNodeEvent)(nil), // 5: iop.EdgeNodeEvent
(*Usage)(nil), // 6: iop.Usage
(*Heartbeat)(nil), // 7: iop.Heartbeat
(*CancelRequest)(nil), // 8: iop.CancelRequest
(*NodeCommandRequest)(nil), // 9: iop.NodeCommandRequest
(*NodeCommandResponse)(nil), // 10: iop.NodeCommandResponse
(*ProviderSnapshot)(nil), // 11: iop.ProviderSnapshot
(*AgentUsageStatus)(nil), // 12: iop.AgentUsageStatus
(*Error)(nil), // 13: iop.Error
(*RegisterRequest)(nil), // 14: iop.RegisterRequest
(*RegisterResponse)(nil), // 15: iop.RegisterResponse
(*NodeConfigPayload)(nil), // 16: iop.NodeConfigPayload
(*AdapterConfig)(nil), // 17: iop.AdapterConfig
(*MockAdapterConfig)(nil), // 18: iop.MockAdapterConfig
(*CLIAdapterConfig)(nil), // 19: iop.CLIAdapterConfig
(*CLIProfileConfig)(nil), // 20: iop.CLIProfileConfig
(*CLICompletionMarker)(nil), // 21: iop.CLICompletionMarker
(*OllamaAdapterConfig)(nil), // 22: iop.OllamaAdapterConfig
(*VllmAdapterConfig)(nil), // 23: iop.VllmAdapterConfig
(*OpenAICompatAdapterConfig)(nil), // 24: iop.OpenAICompatAdapterConfig
(*NodeRuntimeConfig)(nil), // 25: iop.NodeRuntimeConfig
nil, // 26: iop.RunRequest.MetadataEntry
nil, // 27: iop.RunEvent.MetadataEntry
nil, // 28: iop.EdgeNodeEvent.MetadataEntry
nil, // 29: iop.NodeCommandRequest.MetadataEntry
nil, // 30: iop.NodeCommandResponse.ResultEntry
nil, // 31: iop.AgentUsageStatus.MetadataEntry
nil, // 32: iop.CLIAdapterConfig.ProfilesEntry
nil, // 33: iop.OpenAICompatAdapterConfig.HeadersEntry
(*structpb.Struct)(nil), // 34: google.protobuf.Struct
}
var file_proto_iop_runtime_proto_depIdxs = []int32{
32, // 0: iop.RunRequest.policy:type_name -> google.protobuf.Struct
32, // 1: iop.RunRequest.input:type_name -> google.protobuf.Struct
25, // 2: iop.RunRequest.metadata:type_name -> iop.RunRequest.MetadataEntry
34, // 0: iop.RunRequest.policy:type_name -> google.protobuf.Struct
34, // 1: iop.RunRequest.input:type_name -> google.protobuf.Struct
26, // 2: iop.RunRequest.metadata:type_name -> iop.RunRequest.MetadataEntry
0, // 3: iop.RunRequest.session_mode:type_name -> iop.RunSessionMode
6, // 4: iop.RunEvent.usage:type_name -> iop.Usage
26, // 5: iop.RunEvent.metadata:type_name -> iop.RunEvent.MetadataEntry
27, // 6: iop.EdgeNodeEvent.metadata:type_name -> iop.EdgeNodeEvent.MetadataEntry
27, // 5: iop.RunEvent.metadata:type_name -> iop.RunEvent.MetadataEntry
28, // 6: iop.EdgeNodeEvent.metadata:type_name -> iop.EdgeNodeEvent.MetadataEntry
1, // 7: iop.CancelRequest.action:type_name -> iop.CancelAction
2, // 8: iop.NodeCommandRequest.type:type_name -> iop.NodeCommandType
28, // 9: iop.NodeCommandRequest.metadata:type_name -> iop.NodeCommandRequest.MetadataEntry
29, // 9: iop.NodeCommandRequest.metadata:type_name -> iop.NodeCommandRequest.MetadataEntry
2, // 10: iop.NodeCommandResponse.type:type_name -> iop.NodeCommandType
12, // 11: iop.NodeCommandResponse.usage_status:type_name -> iop.AgentUsageStatus
29, // 12: iop.NodeCommandResponse.result:type_name -> iop.NodeCommandResponse.ResultEntry
30, // 12: iop.NodeCommandResponse.result:type_name -> iop.NodeCommandResponse.ResultEntry
11, // 13: iop.NodeCommandResponse.provider_snapshots:type_name -> iop.ProviderSnapshot
30, // 14: iop.AgentUsageStatus.metadata:type_name -> iop.AgentUsageStatus.MetadataEntry
31, // 14: iop.AgentUsageStatus.metadata:type_name -> iop.AgentUsageStatus.MetadataEntry
16, // 15: iop.RegisterResponse.config:type_name -> iop.NodeConfigPayload
17, // 16: iop.NodeConfigPayload.adapters:type_name -> iop.AdapterConfig
24, // 17: iop.NodeConfigPayload.runtime:type_name -> iop.NodeRuntimeConfig
32, // 18: iop.AdapterConfig.settings:type_name -> google.protobuf.Struct
25, // 17: iop.NodeConfigPayload.runtime:type_name -> iop.NodeRuntimeConfig
34, // 18: iop.AdapterConfig.settings:type_name -> google.protobuf.Struct
19, // 19: iop.AdapterConfig.cli:type_name -> iop.CLIAdapterConfig
22, // 20: iop.AdapterConfig.ollama:type_name -> iop.OllamaAdapterConfig
23, // 21: iop.AdapterConfig.vllm:type_name -> iop.VllmAdapterConfig
18, // 22: iop.AdapterConfig.mock:type_name -> iop.MockAdapterConfig
31, // 23: iop.CLIAdapterConfig.profiles:type_name -> iop.CLIAdapterConfig.ProfilesEntry
21, // 24: iop.CLIProfileConfig.completion_marker:type_name -> iop.CLICompletionMarker
20, // 25: iop.CLIAdapterConfig.ProfilesEntry.value:type_name -> iop.CLIProfileConfig
26, // [26:26] is the sub-list for method output_type
26, // [26:26] is the sub-list for method input_type
26, // [26:26] is the sub-list for extension type_name
26, // [26:26] is the sub-list for extension extendee
0, // [0:26] is the sub-list for field type_name
24, // 23: iop.AdapterConfig.openai_compat:type_name -> iop.OpenAICompatAdapterConfig
32, // 24: iop.CLIAdapterConfig.profiles:type_name -> iop.CLIAdapterConfig.ProfilesEntry
21, // 25: iop.CLIProfileConfig.completion_marker:type_name -> iop.CLICompletionMarker
33, // 26: iop.OpenAICompatAdapterConfig.headers:type_name -> iop.OpenAICompatAdapterConfig.HeadersEntry
20, // 27: iop.CLIAdapterConfig.ProfilesEntry.value:type_name -> iop.CLIProfileConfig
28, // [28:28] is the sub-list for method output_type
28, // [28:28] is the sub-list for method input_type
28, // [28:28] is the sub-list for extension type_name
28, // [28:28] is the sub-list for extension extendee
0, // [0:28] is the sub-list for field type_name
}
func init() { file_proto_iop_runtime_proto_init() }
@ -2224,6 +2349,7 @@ func file_proto_iop_runtime_proto_init() {
(*AdapterConfig_Ollama)(nil),
(*AdapterConfig_Vllm)(nil),
(*AdapterConfig_Mock)(nil),
(*AdapterConfig_OpenaiCompat)(nil),
}
type x struct{}
out := protoimpl.TypeBuilder{
@ -2231,7 +2357,7 @@ func file_proto_iop_runtime_proto_init() {
GoPackagePath: reflect.TypeOf(x{}).PkgPath(),
RawDescriptor: unsafe.Slice(unsafe.StringData(file_proto_iop_runtime_proto_rawDesc), len(file_proto_iop_runtime_proto_rawDesc)),
NumEnums: 3,
NumMessages: 29,
NumMessages: 31,
NumExtensions: 0,
NumServices: 0,
},

View file

@ -164,14 +164,15 @@ message NodeConfigPayload {
// adapters it may be empty (equivalent to the type name). When a node carries
// multiple instances of the same adapter type each must have a unique name.
message AdapterConfig {
string type = 1; // "mock" | "ollama" | "vllm" | "cli"
string type = 1; // "mock" | "ollama" | "vllm" | "cli" | "openai_compat"
bool enabled = 2;
google.protobuf.Struct settings = 3; // legacy/compat path; new adapters use oneof
oneof config {
CLIAdapterConfig cli = 4;
OllamaAdapterConfig ollama = 5;
VllmAdapterConfig vllm = 6;
MockAdapterConfig mock = 7;
CLIAdapterConfig cli = 4;
OllamaAdapterConfig ollama = 5;
VllmAdapterConfig vllm = 6;
MockAdapterConfig mock = 7;
OpenAICompatAdapterConfig openai_compat = 10;
}
string name = 8; // stable instance identity; empty = legacy single-instance
string target = 9; // optional default target/route hint for this instance
@ -219,6 +220,16 @@ message VllmAdapterConfig {
int32 request_timeout_ms = 5;
}
message OpenAICompatAdapterConfig {
string provider = 1;
string endpoint = 2;
map<string, string> headers = 3;
int32 capacity = 4;
int32 max_queue = 5;
int32 queue_timeout_ms = 6;
int32 request_timeout_ms = 7;
}
// NodeRuntimeConfig is the runtime tuning pushed to the node.
message NodeRuntimeConfig {
int32 concurrency = 1;