pi-provider-kimi-code 0.4.0 → 0.5.0

package/README.md

@@ -166,9 +166,9 @@ The full env list, including base URL overrides, `kimi-cli` path overrides, uplo
 
 ### Cache behavior
 
-Kimi's cache is content-based. It fires automatically when your prompt prefix matches an earlier request. `prompt_cache_key` and Anthropic `cache_control` markers do not control cache hits on the Coding endpoint.
+Kimi's cache is content-based. It fires automatically when your prompt prefix matches an earlier request. `prompt_cache_key`, Anthropic `cache_control` markers, and identifying headers like `X-Msh-Device-Id` are all ignored for cache decisions — cache is keyed purely by prompt content and is shared across machines, agent forks, and device-id rotations on the same Kimi account.
 
-See [docs/caching.md](docs/caching.md) for the measured behavior: TTL, cross-protocol cache reuse, 256-token prefix alignment, and the cases that invalidate cache.
+See [docs/caching.md](docs/caching.md) for the measured behavior: TTL, cross-protocol cache reuse, 256-token prefix alignment, device-id sharing, and the cases that invalidate cache.
 
 ### Protocol modes
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-provider-kimi-code",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "Use Kimi Code in Pi without losing the Kimi-specific bits",
   "keywords": [
     "pi-package",

package/src/constants.ts

@@ -24,11 +24,17 @@ export const DEFAULT_BASE_URL = IS_OPENAI_PROTOCOL
   ? "https://api.kimi.com/coding/v1"
   : "https://api.kimi.com/coding";
 
-export const KIMI_CLI_VERSION = "1.44.0";
-export const KIMI_CLI_USER_AGENT = `KimiCLI/${KIMI_CLI_VERSION}`;
-export const KIMI_PLATFORM = "kimi_cli";
+export const KIMI_CODE_VERSION = "0.1.1";
+export const KIMI_CODE_USER_AGENT_PRODUCT = "kimi-code-cli";
+export const KIMI_CODE_USER_AGENT = `${KIMI_CODE_USER_AGENT_PRODUCT}/${KIMI_CODE_VERSION}`;
+export const KIMI_PLATFORM = "kimi-code-cli";
+
+export function getKimiCodeHome(): string {
+  const value = process.env.KIMI_CODE_HOME?.trim();
+  return value || join(os.homedir(), ".kimi-code");
+}
 
-export const DEVICE_ID_PATH = join(os.homedir(), ".pi", "providers", "kimi-coding", "device_id");
+export const DEVICE_ID_PATH = join(getKimiCodeHome(), "device_id");
 
 export const DEFAULT_KIMI_MODEL_INPUT = ["text", "image"] as const;
 

package/src/device.ts

@@ -1,8 +1,7 @@
-// Device identification: device-id persistence, kimi-cli-compatible request
+// Device identification: device-id persistence, Kimi Code-compatible request
 // headers, and the cross-platform device-model string. Pure helpers live near
 // the side-effecting one-shots (file IO, exec) they support.
 
-import { execSync } from "node:child_process";
 import { randomBytes } from "node:crypto";
 import { chmodSync, existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import os from "node:os";
@@ -10,8 +9,8 @@ import { dirname } from "node:path";
 
 import {
   DEVICE_ID_PATH,
-  KIMI_CLI_USER_AGENT,
-  KIMI_CLI_VERSION,
+  KIMI_CODE_USER_AGENT,
+  KIMI_CODE_VERSION,
   KIMI_PLATFORM,
 } from "./constants.ts";
 
@@ -47,16 +46,8 @@ function persistDeviceId(deviceId: string): void {
   }
 }
 
-function getMacOSVersion(): string {
-  try {
-    return execSync("sw_vers -productVersion", { encoding: "utf8" }).trim();
-  } catch {
-    return os.release();
-  }
-}
-
 // Normalize Node's lower-case `process.platform` (`linux`, `freebsd`,
-// `sunos`...) to Python `platform.system()` casing used by upstream kimi-cli.
+// `sunos`...) to the canonical OS names used by Kimi Code identity headers.
 const SYSTEM_NAME: Record<string, string> = {
   aix: "AIX",
   freebsd: "FreeBSD",
@@ -73,8 +64,8 @@ export interface DeviceModelInput {
   macVersion?: string;
 }
 
-// Pure function exposed for tests. Mirror of upstream kimi-cli's
-// `_device_model()` in `src/kimi_cli/auth/oauth.py`.
+// Pure function exposed for tests. Mirrors upstream Kimi Code identity
+// formatting in `packages/oauth/src/identity.ts`.
 export function computeDeviceModel(input: DeviceModelInput): string {
   const { platform, release, arch, macVersion } = input;
   if (platform === "darwin") {
@@ -110,12 +101,11 @@ function getDeviceModel(): string {
     platform: process.platform,
     release: os.release(),
     arch: os.machine() || process.arch,
-    macVersion: process.platform === "darwin" ? getMacOSVersion() : undefined,
   });
 }
 
 export function getOsVersion(): string {
-  return os.version();
+  return os.release();
 }
 
 export function asciiHeaderValue(value: string, fallback = "unknown"): string {
@@ -150,9 +140,9 @@ function getStableDeviceId(): string {
 
 export function getCommonHeaders(): Record<string, string> {
   const headers = {
-    "User-Agent": KIMI_CLI_USER_AGENT,
+    "User-Agent": KIMI_CODE_USER_AGENT,
     "X-Msh-Platform": KIMI_PLATFORM,
-    "X-Msh-Version": KIMI_CLI_VERSION,
+    "X-Msh-Version": KIMI_CODE_VERSION,
     "X-Msh-Device-Name": os.hostname(),
     "X-Msh-Device-Model": DEVICE_MODEL,
     "X-Msh-Os-Version": getOsVersion(),

package/src/oauth.ts

@@ -1,4 +1,4 @@
-// OAuth subsystem: device flow, refresh-with-retry, kimi-cli credential
+// OAuth subsystem: device flow, refresh-with-retry, Kimi credential
 // reuse, the login / refresh handlers wired into pi's OAuth interface, and
 // the stream-level auth refresh used by the streaming handler.
 
@@ -183,13 +183,13 @@ export async function refreshAccessToken(
 }
 
 // =============================================================================
-// Reuse existing kimi-cli credentials
+// Reuse existing Kimi credentials
 //
-// Users who already ran the upstream `kimi-cli` and signed in have a valid
-// OAuth token sitting at `$KIMI_SHARE_DIR/credentials/kimi-code.json`
-// (defaults to `~/.kimi/...`). Loading it lets users skip the device-flow
-// dance entirely. Read-only: we never overwrite kimi-cli's file, only seed
-// pi's auth.json via the value returned to pi's OAuth callback.
+// Users who already ran the upstream Kimi Code CLI and signed in have a valid
+// OAuth token under `$KIMI_CODE_HOME/credentials/kimi-code.json` (defaults to
+// `~/.kimi-code/...`). We also keep read-only support for legacy kimi-cli
+// credentials under `$KIMI_SHARE_DIR/credentials/kimi-code.json` (defaults to
+// `~/.kimi/...`). Loading either lets users skip the device-flow dance.
 // =============================================================================
 
 interface KimiCliCredentialsFile {
@@ -198,21 +198,27 @@ interface KimiCliCredentialsFile {
   expires_at?: number; // Unix seconds (upstream convention)
 }
 
-function getKimiCliCredentialsPath(): string {
+function getKimiCredentialPaths(): string[] {
+  const kimiCodeHome = process.env.KIMI_CODE_HOME || join(os.homedir(), ".kimi-code");
   const shareDir = process.env.KIMI_SHARE_DIR || join(os.homedir(), ".kimi");
-  return join(shareDir, "credentials", "kimi-code.json");
+  return [
+    join(kimiCodeHome, "credentials", "kimi-code.json"),
+    join(shareDir, "credentials", "kimi-code.json"),
+  ];
 }
 
 function readKimiCliCredentials(): KimiCliCredentialsFile | null {
-  const path = getKimiCliCredentialsPath();
-  try {
-    if (!existsSync(path)) return null;
-    const data = JSON.parse(readFileSync(path, "utf-8")) as KimiCliCredentialsFile;
-    if (!data.access_token || !data.refresh_token) return null;
-    return data;
-  } catch {
-    return null;
+  for (const path of getKimiCredentialPaths()) {
+    try {
+      if (!existsSync(path)) continue;
+      const data = JSON.parse(readFileSync(path, "utf-8")) as KimiCliCredentialsFile;
+      if (!data.access_token || !data.refresh_token) continue;
+      return data;
+    } catch {
+      continue;
+    }
   }
+  return null;
 }
 
 async function tryReuseKimiCliCredentials(
@@ -226,7 +232,7 @@ async function tryReuseKimiCliCredentials(
   // between this returning and the first API call.
   const stillFresh = expiresAtMs > Date.now() + 60_000;
 
-  callbacks.onProgress?.("Found existing kimi-cli credentials, reusing them.");
+  callbacks.onProgress?.("Found existing Kimi credentials, reusing them.");
 
   if (stillFresh) {
     const extras = await discoverKimiModelMetadata(data.access_token!);
@@ -238,7 +244,7 @@ async function tryReuseKimiCliCredentials(
     };
   }
 
-  callbacks.onProgress?.("kimi-cli access token expired, refreshing.");
+  callbacks.onProgress?.("Kimi access token expired, refreshing.");
   try {
     const token = await refreshAccessToken(data.refresh_token!);
     const extras = await discoverKimiModelMetadata(token.access_token);
@@ -249,7 +255,7 @@ async function tryReuseKimiCliCredentials(
       ...extras,
     };
   } catch {
-    callbacks.onProgress?.("Refresh of kimi-cli token failed, falling back to device flow.");
+    callbacks.onProgress?.("Refresh of Kimi token failed, falling back to device flow.");
     return null;
   }
 }

package/src/payload.ts

@@ -19,7 +19,7 @@ export type Uploader = (mimeType: string, data: string) => Promise<string | null
 export interface KimiEnvOverrides {
   temperature?: number;
   topP?: number;
-  maxTokens?: number;
+  maxCompletionTokens?: number;
 }
 
 export function resolveCacheRetention(value?: CacheRetention): CacheRetention {
@@ -87,8 +87,9 @@ export function readEnvOverrides(): KimiEnvOverrides {
   if (temp) out.temperature = parseFloat(temp);
   const topP = process.env.KIMI_MODEL_TOP_P;
   if (topP) out.topP = parseFloat(topP);
-  const maxTokens = process.env.KIMI_MODEL_MAX_TOKENS;
-  if (maxTokens) out.maxTokens = parseInt(maxTokens, 10);
+  const maxCompletionTokens =
+    process.env.KIMI_MODEL_MAX_COMPLETION_TOKENS ?? process.env.KIMI_MODEL_MAX_TOKENS;
+  if (maxCompletionTokens) out.maxCompletionTokens = parseInt(maxCompletionTokens, 10);
   return out;
 }
 
@@ -421,10 +422,13 @@ export async function applyKimiPayloadMutations(
   }
 
   // 4. Env-level hyperparameter overrides (pre-parsed into numbers by caller).
-  const { temperature, topP, maxTokens } = ctx.envOverrides;
+  const { temperature, topP, maxCompletionTokens } = ctx.envOverrides;
   if (temperature !== undefined) payload.temperature = temperature;
   if (topP !== undefined) payload.top_p = topP;
-  if (maxTokens !== undefined) payload.max_tokens = maxTokens;
+  if (maxCompletionTokens !== undefined) {
+    payload.max_completion_tokens = maxCompletionTokens;
+    delete payload.max_tokens;
+  }
 
   // 5. Reasoning effort mapping.
   if (ctx.reasoning) {