1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
|
import type { ExtensionAPI } from "@mariozechner/pi-coding-agent";
interface ProviderConfig {
id: string;
baseUrl: string;
envKey: string;
suffix: string;
}
const PROVIDERS: ProviderConfig[] = [
{
id: "siliconflow",
baseUrl: "https://api.siliconflow.com/v1",
envKey: "SILICONFLOW_API_KEY",
suffix: "SF"
},
{
id: "deepinfra",
baseUrl: "https://api.deepinfra.com/v1/openai",
envKey: "DEEPINFRA_API_KEY",
suffix: "DI"
}
];
// Patterns matched in order; first match wins.
const CONTEXT_OVERRIDES: Array<[RegExp, number]> = [
[/qwen3.*coder/i, 262144],
[/qwen3.*next/i, 262144],
[/qwen3-235b|qwen3-32b|qwen3-14b|qwen3-8b|qwen3-30b/i, 131072],
[/qwq-32b/i, 131072],
[/qwen2\.5-.*-128k/i, 131072],
[/qwen2\.5/i, 32768],
[/deepseek-v3|deepseek-r1/i, 131072],
[/kimi-k2/i, 131072],
[/glm-4\.[567]|glm-5/i, 131072],
[/gpt-oss/i, 131072],
[/minimax/i, 245760],
[/hunyuan/i, 32768],
];
function resolveContextWindow(modelId: string): number {
for (const [pattern, ctx] of CONTEXT_OVERRIDES) {
if (pattern.test(modelId)) return ctx;
}
return 32768; // conservative fallback
}
export default async function (pi: ExtensionAPI) {
console.log("Provider discovery: Starting...");
await Promise.all(PROVIDERS.map(async (config) => {
const apiKey = process.env[config.envKey];
if (!apiKey) {
console.warn(`Provider discovery: ${config.envKey} not found. Skipping ${config.id}.`);
return;
}
try {
const response = await fetch(`${config.baseUrl}/models`, {
headers: { "Authorization": `Bearer ${apiKey}` }
});
if (!response.ok) {
throw new Error(`API error ${response.status}`);
}
const payload = await response.json();
const models = payload.data
.filter((model: any) => {
// Heuristic to filter for chat/text models
const id = model.id.toLowerCase();
return id.includes("/") || model.type === "text-generation" || id.includes("gpt") || id.includes("llama") || id.includes("claude");
})
.map((model: any) => {
const id = model.id;
const isReasoning = id.toLowerCase().includes("r1") ||
id.toLowerCase().includes("thinking") ||
id.toLowerCase().includes("reasoning") ||
id.toLowerCase().includes("qwq");
return {
id: id,
name: `${id.split('/').pop()} (${config.suffix})`,
reasoning: isReasoning,
input: ["text"],
contextWindow: model.context_window ?? resolveContextWindow(id),
maxTokens: 4096
};
});
pi.registerProvider(config.id, {
baseUrl: config.baseUrl,
apiKey: config.envKey,
api: "openai-completions",
models: models,
compat: {
supportsDeveloperRole: false,
supportsStrictMode: false,
requiresToolResultName: true
},
});
console.log(`Provider discovery: Registered ${models.length} models for ${config.id}.`);
} catch (error) {
console.error(`Provider discovery: Failed to fetch models for ${config.id}:`, error);
}
}));
}
|
