mslxdff 0.1.156 → 0.1.159

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (97) hide show
  1. package/bin/mslxdff.js +4 -4
  2. package/docs/cli_help_mini.md +133 -0
  3. package/package.json +3 -3
  4. package/src/auto.js +254 -254
  5. package/src/bench/cline-bench.js +42 -42
  6. package/src/bench/probe.js +70 -70
  7. package/src/bench/report.js +162 -162
  8. package/src/bench/runner.js +77 -77
  9. package/src/bench/via-probe.js +124 -124
  10. package/src/bench/via-routes.js +87 -87
  11. package/src/bench/workbuddy-bench.js +54 -54
  12. package/src/chat/engine.js +160 -160
  13. package/src/chat/gateway.js +163 -163
  14. package/src/chat/orchestrator.js +234 -234
  15. package/src/chat/prompt.js +70 -70
  16. package/src/chat/repl.js +88 -88
  17. package/src/chat/terminal.js +135 -135
  18. package/src/chat/tools.js +306 -306
  19. package/src/chat-pipeline/index.js +123 -123
  20. package/src/chat-pipeline/policy.js +76 -76
  21. package/src/chat-pipeline/serial-trial.js +210 -210
  22. package/src/cli/commands/group.js +249 -249
  23. package/src/cli/commands/model/list-providers.js +1 -1
  24. package/src/cli/commands/model/picks.js +50 -50
  25. package/src/cli/commands/provider/bench-via.js +247 -247
  26. package/src/cli/commands/provider/bench.js +141 -141
  27. package/src/cli/commands/provider/index.js +124 -116
  28. package/src/cli/commands/provider/models.js +139 -128
  29. package/src/cli/commands/provider/qwenwork-login.js +119 -0
  30. package/src/cli/commands/provider/zcode-login.js +77 -0
  31. package/src/cli/commands/provider/zcode-quota.js +55 -0
  32. package/src/cli/commands/sync.js +232 -232
  33. package/src/cli/provider-row.js +2 -2
  34. package/src/cli/status.js +279 -279
  35. package/src/daemon.js +96 -96
  36. package/src/model-capabilities/enrich.js +86 -86
  37. package/src/model-capabilities/index.js +183 -183
  38. package/src/model-capabilities/parse.js +70 -70
  39. package/src/models.js +225 -225
  40. package/src/providers/classify.js +1 -1
  41. package/src/providers/cline/auth.js +228 -228
  42. package/src/providers/cline/chat.js +307 -307
  43. package/src/providers/cline.js +2 -2
  44. package/src/providers/keyring.js +60 -56
  45. package/src/providers/qoder/chat.js +183 -174
  46. package/src/providers/qoder/index.js +230 -185
  47. package/src/providers/qoder/sse.js +103 -62
  48. package/src/providers/qwenwork/account-store.js +133 -0
  49. package/src/providers/qwenwork/constants.js +67 -0
  50. package/src/providers/qwenwork/cosy.js +120 -0
  51. package/src/providers/qwenwork/crypto.js +218 -0
  52. package/src/providers/qwenwork/http.js +20 -0
  53. package/src/providers/qwenwork/index.js +327 -0
  54. package/src/providers/qwenwork/payload.js +142 -0
  55. package/src/providers/qwenwork/rsa.js +54 -0
  56. package/src/providers/qwenwork/sse.js +268 -0
  57. package/src/providers/qwenwork/stream.js +130 -0
  58. package/src/providers/qwenwork/upstream.js +120 -0
  59. package/src/providers/qwenwork.js +1 -0
  60. package/src/providers/registry.js +66 -56
  61. package/src/providers/share-keys.js +2 -2
  62. package/src/providers/workbuddy/chat.js +248 -248
  63. package/src/providers/workbuddy/reshape.js +152 -152
  64. package/src/providers/workbuddy.js +2 -2
  65. package/src/providers/zcode/account-store.js +129 -0
  66. package/src/providers/zcode/auth.js +28 -0
  67. package/src/providers/zcode/chat.js +171 -0
  68. package/src/providers/zcode/const.js +54 -0
  69. package/src/providers/zcode/headers.js +55 -0
  70. package/src/providers/zcode/index.js +124 -0
  71. package/src/providers/zcode/models.js +65 -0
  72. package/src/providers/zcode/oauth.js +120 -0
  73. package/src/providers/zcode/quota.js +176 -0
  74. package/src/providers/zcode/sse.js +179 -0
  75. package/src/reasoning.js +32 -32
  76. package/src/routes/chat/gateway.js +46 -46
  77. package/src/routes/chat/relay-pipeline.js +250 -250
  78. package/src/routes/chat/via-route-handler.js +144 -144
  79. package/src/routes/hedge.js +255 -255
  80. package/src/routes/models-route.js +167 -167
  81. package/src/routes/peers.js +273 -273
  82. package/src/routes/stream.js +438 -438
  83. package/src/runtime/bootstrap.js +45 -45
  84. package/src/runtime/provider-gate.js +33 -30
  85. package/src/runtime/providers-setup.js +165 -165
  86. package/src/server.js +64 -64
  87. package/src/state/schemas/allowlist.js +92 -92
  88. package/src/sync-opencode.js +280 -280
  89. package/src/transport/index.js +244 -244
  90. package/src/transport/pool.js +56 -56
  91. package/src/transport/retry.js +24 -24
  92. package/src/transport/sse.js +93 -93
  93. package/src/upstream-probe/display.js +52 -52
  94. package/src/upstream-probe/probe.js +49 -49
  95. package/src/upstream-probe/rotate.js +110 -110
  96. package/src/upstream-probe/start.js +45 -45
  97. package/src/upstream.js +289 -289
@@ -1,125 +1,125 @@
1
- import { joinUrl } from "../providers/base.js";
2
- import { computeMetrics, extractUsageFromJson } from "../metrics.js";
3
- import { createTransport } from "../transport/index.js";
1
+ import { joinUrl } from "../providers/base.js";
2
+ import { computeMetrics, extractUsageFromJson } from "../metrics.js";
3
+ import { createTransport } from "../transport/index.js";
4
4
  import { compatFetch } from "../compat.js";
5
-
6
- function extractInnerMessage(bodyText) {
7
- const t = String(bodyText || "");
8
- try {
9
- const j = JSON.parse(t);
10
- const m = j?.error?.message || j?.error || j?.message || j?.data?.error || "";
11
- if (typeof m === "string" && m.trim()) return m.trim().slice(0, 300);
12
- if (typeof j?.error === "string") return j.error.slice(0, 300);
13
- } catch {}
14
- return t.slice(0, 300);
15
- }
16
-
17
- function classifyError(status, bodyText) {
18
- const t = String(bodyText || "").slice(0, 500);
19
- const low = t.toLowerCase();
20
- if (status === 401) return { label: "鉴权失败", retryable: false };
21
- if (status === 402 || /insufficient balance/i.test(t)) return { label: "余额不足", retryable: false };
22
- if (low.includes("only available via cline")) return { label: "仅 Cline 客户端可用", retryable: false };
23
- if (low.includes("invalid model format")) return { label: "模型格式错误", retryable: false };
24
- if (status === 403) return { label: /insufficient/i.test(t) ? "余额不足" : "鉴权失败", retryable: false };
25
- if (status === 429) return { label: "限流", retryable: true };
26
- if (status >= 500) return { label: `上游错误 ${status}`, retryable: true };
27
- if (status === 404) return { label: "模型不存在", retryable: false };
28
- return { label: `HTTP ${status}`, retryable: false };
29
- }
30
-
31
- export async function viaProbe({
32
- peerUrl,
33
- token,
34
- providerId,
35
- model,
36
- prompt = "hi",
37
- maxTokens = 5,
38
- timeoutMs = 30000,
39
- fetchImpl = compatFetch,
40
- clock = Date.now,
41
- shareKeys,
42
- shareKeysHeader,
43
- relayTarget,
44
- relayHeaders,
45
- relayBody,
46
- targetUrl,
47
- } = {}) {
48
- const base = String(peerUrl || "").replace(/\/+$/, "");
49
- if (!base) return { ok: false, label: "配置错误", error: "missing peerUrl", ttfbMs: null, totalMs: 0, tps: null, charsPerSec: null, tokens: null };
50
- const tr = createTransport({ fetchImpl, keepAlive: false, retry: {}, timeoutMs });
51
- const rt = String(relayTarget || targetUrl || "").trim();
52
- if (rt) {
53
- const rh = relayHeaders && typeof relayHeaders === "object" ? relayHeaders : {};
54
- const rb = relayBody !== undefined ? relayBody : null;
55
- const relayUrl = joinUrl(base, "/v1/relay");
56
- const relayHeadersOut = { "Content-Type": "application/json", Accept: "application/json" };
57
- if (token) relayHeadersOut.Authorization = `Bearer ${token}`;
58
- const payload = { targetUrl: rt, method: "POST", headers: rh, body: rb };
59
- try {
60
- const res = await tr.request({ url: relayUrl, method: "POST", headers: relayHeadersOut, body: payload, stream: false });
61
- const ttfbMs = res.ttfbMs;
62
- const totalMs = res.totalMs;
63
- if (!res.ok) {
64
- let txt = ""; try { txt = await res.text(); } catch {}
65
- const cls = classifyError(res.status, txt);
66
- const msg = extractInnerMessage(txt) || `HTTP ${res.status}`;
67
- return { ok: false, status: res.status, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
68
- }
69
- let txt = ""; let j = {};
70
- try { txt = await res.text(); j = JSON.parse(txt); } catch { j = {}; }
71
- const relayStatus = res.headers.get("x-mslxdff-relay-status") ? Number(res.headers.get("x-mslxdff-relay-status")) : res.status;
72
- if (relayStatus >= 400) {
73
- const cls = classifyError(relayStatus, txt);
74
- const msg = extractInnerMessage(txt) || `HTTP ${relayStatus}`;
75
- return { ok: false, status: relayStatus, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
76
- }
77
- const usage = extractUsageFromJson(j);
78
- const content = j?.choices?.[0]?.message?.content || j?.choices?.[0]?.text || txt || "";
79
- const chars = typeof content === "string" ? content.length : 0;
80
- const pt = usage?.prompt_tokens ?? null;
81
- const ct = usage?.completion_tokens ?? null;
82
- const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens: pt, completionTokens: ct, chars });
83
- const totalTokens = usage?.total_tokens ?? (pt !== null && ct !== null ? pt + ct : null);
84
- return { ok: true, status: relayStatus, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: { prompt: pt, completion: ct, total: totalTokens }, chars };
85
- } catch (e) {
86
- const msg = e?.message || String(e);
87
- const isTimeout = /timeout|abort/i.test(msg);
88
- return { ok: false, label: isTimeout ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs: null, totalMs: null, tps: null, charsPerSec: null, tokens: null };
89
- }
90
- }
91
- if (!model) return { ok: false, label: "配置错误", error: "missing model", ttfbMs: null, totalMs: 0, tps: null, charsPerSec: null, tokens: null };
92
- let rawModel = String(model).trim();
93
- if (providerId && rawModel.startsWith(`${providerId}/`)) rawModel = rawModel.slice(providerId.length + 1);
94
- const url = joinUrl(base, "/v1/chat/completions");
95
- const body = { model: rawModel, stream: false, messages: [{ role: "user", content: prompt }], max_tokens: maxTokens };
96
- const headers = { "Content-Type": "application/json", Accept: "application/json" };
97
- if (token) headers.Authorization = `Bearer ${token}`;
98
- const sk = shareKeysHeader || shareKeys;
99
- if (sk) headers["x-mslxdff-share-keys"] = String(sk);
100
- try {
101
- const res = await tr.request({ url, method: "POST", headers, body, stream: false });
102
- const ttfbMs = res.ttfbMs;
103
- const totalMs = res.totalMs;
104
- if (!res.ok) {
105
- let txt = ""; try { txt = await res.text(); } catch {}
106
- const cls = classifyError(res.status, txt);
107
- const msg = extractInnerMessage(txt) || `HTTP ${res.status}`;
108
- return { ok: false, status: res.status, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
109
- }
110
- let json = {}; let txt = "";
111
- try { txt = await res.text(); json = JSON.parse(txt); } catch { json = {}; }
112
- const usage = extractUsageFromJson(json);
113
- const content = json?.choices?.[0]?.message?.content || json?.choices?.[0]?.text || txt || "";
114
- const chars = typeof content === "string" ? content.length : 0;
115
- const promptTokens = usage?.prompt_tokens ?? null;
116
- const completionTokens = usage?.completion_tokens ?? null;
117
- const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens, completionTokens, chars });
118
- const totalTokens = usage?.total_tokens ?? (promptTokens !== null && completionTokens !== null ? promptTokens + completionTokens : null);
119
- return { ok: true, status: res.status, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: { prompt: promptTokens, completion: completionTokens, total: totalTokens }, chars };
120
- } catch (e) {
121
- const msg = e?.message || String(e);
122
- const isTimeout = /timeout|abort/i.test(msg);
123
- return { ok: false, label: isTimeout ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs: null, totalMs: null, tps: null, charsPerSec: null, tokens: null };
124
- }
125
- }
5
+
6
+ function extractInnerMessage(bodyText) {
7
+ const t = String(bodyText || "");
8
+ try {
9
+ const j = JSON.parse(t);
10
+ const m = j?.error?.message || j?.error || j?.message || j?.data?.error || "";
11
+ if (typeof m === "string" && m.trim()) return m.trim().slice(0, 300);
12
+ if (typeof j?.error === "string") return j.error.slice(0, 300);
13
+ } catch {}
14
+ return t.slice(0, 300);
15
+ }
16
+
17
+ function classifyError(status, bodyText) {
18
+ const t = String(bodyText || "").slice(0, 500);
19
+ const low = t.toLowerCase();
20
+ if (status === 401) return { label: "鉴权失败", retryable: false };
21
+ if (status === 402 || /insufficient balance/i.test(t)) return { label: "余额不足", retryable: false };
22
+ if (low.includes("only available via cline")) return { label: "仅 Cline 客户端可用", retryable: false };
23
+ if (low.includes("invalid model format")) return { label: "模型格式错误", retryable: false };
24
+ if (status === 403) return { label: /insufficient/i.test(t) ? "余额不足" : "鉴权失败", retryable: false };
25
+ if (status === 429) return { label: "限流", retryable: true };
26
+ if (status >= 500) return { label: `上游错误 ${status}`, retryable: true };
27
+ if (status === 404) return { label: "模型不存在", retryable: false };
28
+ return { label: `HTTP ${status}`, retryable: false };
29
+ }
30
+
31
+ export async function viaProbe({
32
+ peerUrl,
33
+ token,
34
+ providerId,
35
+ model,
36
+ prompt = "hi",
37
+ maxTokens = 5,
38
+ timeoutMs = 30000,
39
+ fetchImpl = compatFetch,
40
+ clock = Date.now,
41
+ shareKeys,
42
+ shareKeysHeader,
43
+ relayTarget,
44
+ relayHeaders,
45
+ relayBody,
46
+ targetUrl,
47
+ } = {}) {
48
+ const base = String(peerUrl || "").replace(/\/+$/, "");
49
+ if (!base) return { ok: false, label: "配置错误", error: "missing peerUrl", ttfbMs: null, totalMs: 0, tps: null, charsPerSec: null, tokens: null };
50
+ const tr = createTransport({ fetchImpl, keepAlive: false, retry: {}, timeoutMs });
51
+ const rt = String(relayTarget || targetUrl || "").trim();
52
+ if (rt) {
53
+ const rh = relayHeaders && typeof relayHeaders === "object" ? relayHeaders : {};
54
+ const rb = relayBody !== undefined ? relayBody : null;
55
+ const relayUrl = joinUrl(base, "/v1/relay");
56
+ const relayHeadersOut = { "Content-Type": "application/json", Accept: "application/json" };
57
+ if (token) relayHeadersOut.Authorization = `Bearer ${token}`;
58
+ const payload = { targetUrl: rt, method: "POST", headers: rh, body: rb };
59
+ try {
60
+ const res = await tr.request({ url: relayUrl, method: "POST", headers: relayHeadersOut, body: payload, stream: false });
61
+ const ttfbMs = res.ttfbMs;
62
+ const totalMs = res.totalMs;
63
+ if (!res.ok) {
64
+ let txt = ""; try { txt = await res.text(); } catch {}
65
+ const cls = classifyError(res.status, txt);
66
+ const msg = extractInnerMessage(txt) || `HTTP ${res.status}`;
67
+ return { ok: false, status: res.status, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
68
+ }
69
+ let txt = ""; let j = {};
70
+ try { txt = await res.text(); j = JSON.parse(txt); } catch { j = {}; }
71
+ const relayStatus = res.headers.get("x-mslxdff-relay-status") ? Number(res.headers.get("x-mslxdff-relay-status")) : res.status;
72
+ if (relayStatus >= 400) {
73
+ const cls = classifyError(relayStatus, txt);
74
+ const msg = extractInnerMessage(txt) || `HTTP ${relayStatus}`;
75
+ return { ok: false, status: relayStatus, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
76
+ }
77
+ const usage = extractUsageFromJson(j);
78
+ const content = j?.choices?.[0]?.message?.content || j?.choices?.[0]?.text || txt || "";
79
+ const chars = typeof content === "string" ? content.length : 0;
80
+ const pt = usage?.prompt_tokens ?? null;
81
+ const ct = usage?.completion_tokens ?? null;
82
+ const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens: pt, completionTokens: ct, chars });
83
+ const totalTokens = usage?.total_tokens ?? (pt !== null && ct !== null ? pt + ct : null);
84
+ return { ok: true, status: relayStatus, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: { prompt: pt, completion: ct, total: totalTokens }, chars };
85
+ } catch (e) {
86
+ const msg = e?.message || String(e);
87
+ const isTimeout = /timeout|abort/i.test(msg);
88
+ return { ok: false, label: isTimeout ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs: null, totalMs: null, tps: null, charsPerSec: null, tokens: null };
89
+ }
90
+ }
91
+ if (!model) return { ok: false, label: "配置错误", error: "missing model", ttfbMs: null, totalMs: 0, tps: null, charsPerSec: null, tokens: null };
92
+ let rawModel = String(model).trim();
93
+ if (providerId && rawModel.startsWith(`${providerId}/`)) rawModel = rawModel.slice(providerId.length + 1);
94
+ const url = joinUrl(base, "/v1/chat/completions");
95
+ const body = { model: rawModel, stream: false, messages: [{ role: "user", content: prompt }], max_tokens: maxTokens };
96
+ const headers = { "Content-Type": "application/json", Accept: "application/json" };
97
+ if (token) headers.Authorization = `Bearer ${token}`;
98
+ const sk = shareKeysHeader || shareKeys;
99
+ if (sk) headers["x-mslxdff-share-keys"] = String(sk);
100
+ try {
101
+ const res = await tr.request({ url, method: "POST", headers, body, stream: false });
102
+ const ttfbMs = res.ttfbMs;
103
+ const totalMs = res.totalMs;
104
+ if (!res.ok) {
105
+ let txt = ""; try { txt = await res.text(); } catch {}
106
+ const cls = classifyError(res.status, txt);
107
+ const msg = extractInnerMessage(txt) || `HTTP ${res.status}`;
108
+ return { ok: false, status: res.status, label: cls.label, error: msg, ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
109
+ }
110
+ let json = {}; let txt = "";
111
+ try { txt = await res.text(); json = JSON.parse(txt); } catch { json = {}; }
112
+ const usage = extractUsageFromJson(json);
113
+ const content = json?.choices?.[0]?.message?.content || json?.choices?.[0]?.text || txt || "";
114
+ const chars = typeof content === "string" ? content.length : 0;
115
+ const promptTokens = usage?.prompt_tokens ?? null;
116
+ const completionTokens = usage?.completion_tokens ?? null;
117
+ const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens, completionTokens, chars });
118
+ const totalTokens = usage?.total_tokens ?? (promptTokens !== null && completionTokens !== null ? promptTokens + completionTokens : null);
119
+ return { ok: true, status: res.status, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: { prompt: promptTokens, completion: completionTokens, total: totalTokens }, chars };
120
+ } catch (e) {
121
+ const msg = e?.message || String(e);
122
+ const isTimeout = /timeout|abort/i.test(msg);
123
+ return { ok: false, label: isTimeout ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs: null, totalMs: null, tps: null, charsPerSec: null, tokens: null };
124
+ }
125
+ }
@@ -1,87 +1,87 @@
1
- import { readFileSync, existsSync } from "node:fs";
2
- import { join, dirname } from "node:path";
3
- import os from "node:os";
4
- import { defaultStateFile } from "../state/store.js";
5
- import { atomicWriteSync } from "../state/persist.js";
6
-
7
- export function defaultViaRoutesFile() {
8
- if (process.env.MSLXDFF_VIA_ROUTES_FILE) return String(process.env.MSLXDFF_VIA_ROUTES_FILE).trim();
9
- const sf = defaultStateFile();
10
- return join(dirname(sf), "via-routes.json");
11
- }
12
-
13
- function viaTtlMs() {
14
- const raw = process.env.MSLXDFF_VIA_ROUTE_TTL_MS;
15
- if (raw === undefined || raw === null || raw === "") return 0;
16
- const s = String(raw).trim().toLowerCase();
17
- if (s === "0" || s === "off" || s === "false" || s === "no" || s === "disable" || s === "disabled") return 0;
18
- const n = Number(s);
19
- if (Number.isInteger(n) && n >= 0) return n;
20
- return 0;
21
- }
22
-
23
- export function loadViaRoutes(file) {
24
- const f = file || defaultViaRoutesFile();
25
- try {
26
- if (!existsSync(f)) return { version: 1, at: null, routes: {}, meta: {} };
27
- const j = JSON.parse(readFileSync(f, "utf8"));
28
- if (j && typeof j === "object" && j.routes && typeof j.routes === "object") return j;
29
- if (j && typeof j === "object" && !j.routes) return { version: 1, at: j.at || null, routes: j, meta: {} };
30
- return { version: 1, at: null, routes: {}, meta: {} };
31
- } catch {
32
- return { version: 1, at: null, routes: {}, meta: {} };
33
- }
34
- }
35
-
36
- export function getViaRoute(model, { file, ttlMs } = {}) {
37
- const id = String(model || "").trim();
38
- if (!id) return null;
39
- const f = file || defaultViaRoutesFile();
40
- const data = loadViaRoutes(f);
41
- const entry = data.routes?.[id];
42
- if (!entry) return null;
43
- const t = ttlMs !== undefined ? ttlMs : viaTtlMs();
44
- if (t > 0 && entry.at) {
45
- const atMs = Date.parse(entry.at);
46
- if (Number.isFinite(atMs) && Date.now() - atMs > t) return null;
47
- }
48
- return entry;
49
- }
50
-
51
- export function saveViaRoutes(results, { file, meta } = {}) {
52
- const f = file || defaultViaRoutesFile();
53
- const now = new Date().toISOString();
54
- const prev = loadViaRoutes(f);
55
- const nextRoutes = { ...(prev.routes || {}) };
56
- for (const r of results || []) {
57
- const id = String(r.model || r.id || "").trim();
58
- if (!id) continue;
59
- const best = String(r.best || "direct").trim() || "direct";
60
- const direct = r.direct ? { ok: Boolean(r.direct.ok), ttfbMs: r.direct.ttfbMs ?? r.direct.totalMs ?? null, totalMs: r.direct.totalMs ?? null, label: r.direct.label || null, error: r.direct.error ? String(r.direct.error).slice(0, 300) : null } : null;
61
- const via = {};
62
- for (const [k, v] of Object.entries(r.via || {})) {
63
- via[k] = v?.ok ? { ok: true, ttfbMs: v.ttfbMs ?? v.totalMs ?? null, totalMs: v.totalMs ?? null } : { ok: false, ttfbMs: v?.ttfbMs ?? null, totalMs: v?.totalMs ?? null, label: v?.label || v?.error || "offline" };
64
- }
65
- nextRoutes[id] = {
66
- best,
67
- direct,
68
- via,
69
- deltaMs: r.deltaMs ?? null,
70
- provider: r.provider || id.split("/")[0] || "",
71
- at: now,
72
- };
73
- }
74
- const out = {
75
- version: 1,
76
- at: now,
77
- routes: nextRoutes,
78
- meta: meta || prev.meta || {},
79
- };
80
- atomicWriteSync(f, out);
81
- return out;
82
- }
83
-
84
- export function clearViaRoutes(file) {
85
- const f = file || defaultViaRoutesFile();
86
- atomicWriteSync(f, { version: 1, at: new Date().toISOString(), routes: {}, meta: {} });
87
- }
1
+ import { readFileSync, existsSync } from "node:fs";
2
+ import { join, dirname } from "node:path";
3
+ import os from "node:os";
4
+ import { defaultStateFile } from "../state/store.js";
5
+ import { atomicWriteSync } from "../state/persist.js";
6
+
7
+ export function defaultViaRoutesFile() {
8
+ if (process.env.MSLXDFF_VIA_ROUTES_FILE) return String(process.env.MSLXDFF_VIA_ROUTES_FILE).trim();
9
+ const sf = defaultStateFile();
10
+ return join(dirname(sf), "via-routes.json");
11
+ }
12
+
13
+ function viaTtlMs() {
14
+ const raw = process.env.MSLXDFF_VIA_ROUTE_TTL_MS;
15
+ if (raw === undefined || raw === null || raw === "") return 0;
16
+ const s = String(raw).trim().toLowerCase();
17
+ if (s === "0" || s === "off" || s === "false" || s === "no" || s === "disable" || s === "disabled") return 0;
18
+ const n = Number(s);
19
+ if (Number.isInteger(n) && n >= 0) return n;
20
+ return 0;
21
+ }
22
+
23
+ export function loadViaRoutes(file) {
24
+ const f = file || defaultViaRoutesFile();
25
+ try {
26
+ if (!existsSync(f)) return { version: 1, at: null, routes: {}, meta: {} };
27
+ const j = JSON.parse(readFileSync(f, "utf8"));
28
+ if (j && typeof j === "object" && j.routes && typeof j.routes === "object") return j;
29
+ if (j && typeof j === "object" && !j.routes) return { version: 1, at: j.at || null, routes: j, meta: {} };
30
+ return { version: 1, at: null, routes: {}, meta: {} };
31
+ } catch {
32
+ return { version: 1, at: null, routes: {}, meta: {} };
33
+ }
34
+ }
35
+
36
+ export function getViaRoute(model, { file, ttlMs } = {}) {
37
+ const id = String(model || "").trim();
38
+ if (!id) return null;
39
+ const f = file || defaultViaRoutesFile();
40
+ const data = loadViaRoutes(f);
41
+ const entry = data.routes?.[id];
42
+ if (!entry) return null;
43
+ const t = ttlMs !== undefined ? ttlMs : viaTtlMs();
44
+ if (t > 0 && entry.at) {
45
+ const atMs = Date.parse(entry.at);
46
+ if (Number.isFinite(atMs) && Date.now() - atMs > t) return null;
47
+ }
48
+ return entry;
49
+ }
50
+
51
+ export function saveViaRoutes(results, { file, meta } = {}) {
52
+ const f = file || defaultViaRoutesFile();
53
+ const now = new Date().toISOString();
54
+ const prev = loadViaRoutes(f);
55
+ const nextRoutes = { ...(prev.routes || {}) };
56
+ for (const r of results || []) {
57
+ const id = String(r.model || r.id || "").trim();
58
+ if (!id) continue;
59
+ const best = String(r.best || "direct").trim() || "direct";
60
+ const direct = r.direct ? { ok: Boolean(r.direct.ok), ttfbMs: r.direct.ttfbMs ?? r.direct.totalMs ?? null, totalMs: r.direct.totalMs ?? null, label: r.direct.label || null, error: r.direct.error ? String(r.direct.error).slice(0, 300) : null } : null;
61
+ const via = {};
62
+ for (const [k, v] of Object.entries(r.via || {})) {
63
+ via[k] = v?.ok ? { ok: true, ttfbMs: v.ttfbMs ?? v.totalMs ?? null, totalMs: v.totalMs ?? null } : { ok: false, ttfbMs: v?.ttfbMs ?? null, totalMs: v?.totalMs ?? null, label: v?.label || v?.error || "offline" };
64
+ }
65
+ nextRoutes[id] = {
66
+ best,
67
+ direct,
68
+ via,
69
+ deltaMs: r.deltaMs ?? null,
70
+ provider: r.provider || id.split("/")[0] || "",
71
+ at: now,
72
+ };
73
+ }
74
+ const out = {
75
+ version: 1,
76
+ at: now,
77
+ routes: nextRoutes,
78
+ meta: meta || prev.meta || {},
79
+ };
80
+ atomicWriteSync(f, out);
81
+ return out;
82
+ }
83
+
84
+ export function clearViaRoutes(file) {
85
+ const f = file || defaultViaRoutesFile();
86
+ atomicWriteSync(f, { version: 1, at: new Date().toISOString(), routes: {}, meta: {} });
87
+ }
@@ -1,55 +1,55 @@
1
- import { computeMetrics } from "../metrics.js";
2
- import { createTransport } from "../transport/index.js";
1
+ import { computeMetrics } from "../metrics.js";
2
+ import { createTransport } from "../transport/index.js";
3
3
  import { compatFetch } from "../compat.js";
4
-
5
- function buildWorkbuddyHeaders(apiKey, auth) {
6
- const h = {
7
- "Content-Type": "application/json",
8
- Accept: "text/event-stream",
9
- "User-Agent": "CLI/2.115.0 WorkBuddy/2.115.0",
10
- Origin: "https://www.codebuddy.cn",
11
- Referer: "https://www.codebuddy.cn/",
12
- "X-Product": "SaaS",
13
- };
14
- if (apiKey) h["Authorization"] = `Bearer ${apiKey}`;
15
- if (auth?.uid) h["X-User-Id"] = auth.uid;
16
- h["X-Domain"] = auth?.domain || "www.codebuddy.cn";
17
- if (auth?.enterpriseId) { h["X-Enterprise-Id"] = auth.enterpriseId; h["X-Tenant-Id"] = auth.enterpriseId; }
18
- return h;
19
- }
20
-
21
- function sseContent(obj) {
22
- const c = obj?.choices?.[0]?.delta?.content || obj?.choices?.[0]?.message?.content || "";
23
- return typeof c === "string" ? c : "";
24
- }
25
-
26
- export async function workbuddyBenchOne({ baseUrl, chatPath = "/v2/chat/completions", model, apiKey, auth, prompt = "hi", maxTokens = 5, timeoutMs = 30000, fetchImpl = compatFetch }) {
27
- const tr = createTransport({ fetchImpl, keepAlive: false, retry: {}, timeoutMs });
28
- let ttfbMs = null;
29
- let totalMs = null;
30
- let content = "";
31
- try {
32
- const url = String(baseUrl).replace(/\/+$/, "") + String(chatPath || "/v2/chat/completions");
33
- const headers = buildWorkbuddyHeaders(apiKey, auth);
34
- const rawModel = String(model || "").trim();
35
- const body = { model: rawModel, stream: true, messages: [{ role: "user", content: prompt }], max_tokens: maxTokens };
36
- const res = await tr.request({ url, method: "POST", headers, body, stream: true });
37
- if (!res.ok) {
38
- let txt = "";
39
- try { txt = await res.text(); } catch {}
40
- const label = res.status === 401 ? "鉴权失败" : res.status === 402 ? "余额不足" : res.status === 429 ? "限流" : res.status >= 500 ? `上游错误 ${res.status}` : `HTTP ${res.status}`;
41
- return { id: model, ok: false, status: res.status, label, error: txt.slice(0, 300), ttfbMs, totalMs: res.totalMs, tps: null, charsPerSec: null, tokens: null };
42
- }
43
- for await (const ev of res.stream()) {
44
- try { content += sseContent(JSON.parse(ev)); } catch {}
45
- }
46
- ttfbMs = res.ttfbMs;
47
- totalMs = res.totalMs;
48
- const chars = content.length;
49
- const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens: null, completionTokens: null, chars });
50
- return { id: model, ok: true, status: 200, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: null, chars };
51
- } catch (e) {
52
- const msg = e?.message || String(e);
53
- return { id: model, ok: false, label: /timeout|abort/i.test(msg) ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
54
- }
55
- }
4
+
5
+ function buildWorkbuddyHeaders(apiKey, auth) {
6
+ const h = {
7
+ "Content-Type": "application/json",
8
+ Accept: "text/event-stream",
9
+ "User-Agent": "CLI/2.115.0 WorkBuddy/2.115.0",
10
+ Origin: "https://www.codebuddy.cn",
11
+ Referer: "https://www.codebuddy.cn/",
12
+ "X-Product": "SaaS",
13
+ };
14
+ if (apiKey) h["Authorization"] = `Bearer ${apiKey}`;
15
+ if (auth?.uid) h["X-User-Id"] = auth.uid;
16
+ h["X-Domain"] = auth?.domain || "www.codebuddy.cn";
17
+ if (auth?.enterpriseId) { h["X-Enterprise-Id"] = auth.enterpriseId; h["X-Tenant-Id"] = auth.enterpriseId; }
18
+ return h;
19
+ }
20
+
21
+ function sseContent(obj) {
22
+ const c = obj?.choices?.[0]?.delta?.content || obj?.choices?.[0]?.message?.content || "";
23
+ return typeof c === "string" ? c : "";
24
+ }
25
+
26
+ export async function workbuddyBenchOne({ baseUrl, chatPath = "/v2/chat/completions", model, apiKey, auth, prompt = "hi", maxTokens = 5, timeoutMs = 30000, fetchImpl = compatFetch }) {
27
+ const tr = createTransport({ fetchImpl, keepAlive: false, retry: {}, timeoutMs });
28
+ let ttfbMs = null;
29
+ let totalMs = null;
30
+ let content = "";
31
+ try {
32
+ const url = String(baseUrl).replace(/\/+$/, "") + String(chatPath || "/v2/chat/completions");
33
+ const headers = buildWorkbuddyHeaders(apiKey, auth);
34
+ const rawModel = String(model || "").trim();
35
+ const body = { model: rawModel, stream: true, messages: [{ role: "user", content: prompt }], max_tokens: maxTokens };
36
+ const res = await tr.request({ url, method: "POST", headers, body, stream: true });
37
+ if (!res.ok) {
38
+ let txt = "";
39
+ try { txt = await res.text(); } catch {}
40
+ const label = res.status === 401 ? "鉴权失败" : res.status === 402 ? "余额不足" : res.status === 429 ? "限流" : res.status >= 500 ? `上游错误 ${res.status}` : `HTTP ${res.status}`;
41
+ return { id: model, ok: false, status: res.status, label, error: txt.slice(0, 300), ttfbMs, totalMs: res.totalMs, tps: null, charsPerSec: null, tokens: null };
42
+ }
43
+ for await (const ev of res.stream()) {
44
+ try { content += sseContent(JSON.parse(ev)); } catch {}
45
+ }
46
+ ttfbMs = res.ttfbMs;
47
+ totalMs = res.totalMs;
48
+ const chars = content.length;
49
+ const { tps, charsPerSec } = computeMetrics({ ttfbMs, totalMs, promptTokens: null, completionTokens: null, chars });
50
+ return { id: model, ok: true, status: 200, label: "成功", ttfbMs, totalMs, tps, charsPerSec, tokens: null, chars };
51
+ } catch (e) {
52
+ const msg = e?.message || String(e);
53
+ return { id: model, ok: false, label: /timeout|abort/i.test(msg) ? "超时" : "网络错误", error: msg.slice(0, 300), ttfbMs, totalMs, tps: null, charsPerSec: null, tokens: null };
54
+ }
55
+ }