@magpie-community/opencode-workbuddy-auth 0.2.6 → 0.2.8

package/README.md

@@ -286,6 +286,18 @@ client sends it: WorkBuddy doesn't think when a chat names none, so
 OpenCode's default variant got answers without thinking. An effort the
 client names, `none` too, goes as it came.
 
+One value differs from the config, measured with real requests:
+DeepSeek-V4.1-Flash (and its paid `-sg` id) lists `none`, though the
+config says it only reasons: WorkBuddy answers `reasoning_effort: "none"`
+on it with no thinking at all (since 0.2.7, #563).
+
+A model's context is the config's `maxInputTokens` on both sites. 0.2.7
+capped DeepSeek-V4.1-Flash at 100000 on `workbuddy` (copilot.tencent.com)
+from one report's error; China accounts took prompts up to 992,389 tokens
+on it, so 0.2.8 lifted the cap (#561). A prompt WorkBuddy does find too
+long ("prompt is too long: … > … maximum") is answered with its own 400,
+passed on as it came, which magpie takes as the context overflowing.
+
 A chat goes gzipped (`Content-Encoding: gzip`), as WorkBuddy's own client
 sends its model requests (`RequestBodyGzip` in both builds' product.json).
 Claude's Messages route goes as it came, and so does anything that isn't a

package/index.mjs

@@ -28,7 +28,7 @@ const CN_MODELS = [
   ["hy4-preview-f", "Hy4 preview", 1000000, ["high"]],
   ["hy3", "Hy3", 192000, ["low", "high"]],
   ["hy3-x", "Hy3-X", 192000, ["low", "high"]],
-  ["deepseek-v4.1-flash", "Deepseek-V4.1-Flash", 1000000],
+  ["deepseek-v4.1-flash", "Deepseek-V4.1-Flash", 1000000, ["none", "low", "high", "max"]],
   ["glm-5.3", "GLM-5.3", 1000000, ["low", "high", "max"]],
   ["glm-5.3-flash", "GLM-5.3-Flash", 1000000, ["low", "high", "max"]],
   ["glm-5.2", "GLM-5.2", 1000000, ["high", "xhigh"]],
@@ -109,6 +109,28 @@ const SITES = {
   },
 }
 
+// A model's context is its config's maxInputTokens, on both sites. 0.2.7
+// capped deepseek-v4.1-flash at 100000 on copilot.tencent.com from one
+// report's error ("prompt is too long: 100001 tokens > 100000 maximum"),
+// and it was wrong: on 2026-10-31 China Free-plan accounts of three other
+// users took 110,041 to 992,389-token prompts on it there, the whole prompt
+// read (needles recalled), and the vendor's own refusal of 1,157,831 named
+// 1048576 as the maximum (plugins#561). The cap made every client compact at
+// 100k. Where a prompt really is too long, WorkBuddy's 400 says "prompt is
+// too long" and goes to the client as it came (explained), which magpie
+// reads as the context overflowing (context_length_exceeded): the agent is
+// told to compact, and the account isn't rested for it.
+
+// OFF_TAKEN are models that turn thinking off for reasoning_effort "none"
+// though the config says onlyReasoning (so WorkBuddy's own client offers no
+// off switch): deepseek-v4.1-flash, and the same model under its paid id.
+// Measured 2026-10-31 on www.workbuddy.ai: "none" answered 200 with 0
+// reasoning tokens on both, where "low" thought 1221 and 1127; and on the
+// China build by the reporter of plugins#563 (0 reasoning tokens, a 30-item
+// prompt in 14s instead of 50s). Without the level magpie sends "low" for
+// a client's "none".
+const OFF_TAKEN = new Set(["deepseek-v4.1-flash", "deepseek-v4.1-flash-sg"])
+
 // textOnly says whether site drops the images sent to model, whatever its
 // config says. On 2026-10-13, with a real WorkBuddy AI account (Yirasumi on
 // Discord), www.workbuddy.ai's /v3/config marked every model
@@ -3132,7 +3154,8 @@ async function liveModels(site, a) {
         const es = effortsOf(d, site, id)
         if (es.length) {
           const canOff = d.reasoning?.canDisableThinking
-          m.efforts = [...(!d.onlyReasoning && (canOff == null || canOff) ? ["none"] : []), ...es]
+          const off = (!d.onlyReasoning && (canOff == null || canOff)) || OFF_TAKEN.has(id)
+          m.efforts = [...(off ? ["none"] : []), ...es]
           const def = defaultEffortOf(d, es)
           if (def) m.effort = def
         }
@@ -3759,4 +3782,4 @@ export const WorkBuddyAuthPlugin = makePlugin(SITES.workbuddy)
 export const WorkBuddyAIAuthPlugin = makePlugin(SITES["workbuddy-ai"])
 
 // for tests
-export const _internal = { idFor, replyAfterToolsOn, withDefaultEffort, defaultEffortOf, gzipped, FREE_ONLY, PRICE_TTL, keyAccount, remoteConfig, cycleOf, typeUnions, systemAsUser, systemKey, DEFAULT_SYSTEM, EDGE_READS, fitShape, replyAfterTools, AFTER_TOOLS_REFUSED, enterpriseUsageOf, accountName, notDesktop, textOnly, runsOutOf, packagesOf, endOf, enterpriseHeaders, enterpriseOf, liveModels, browserSignIn, withSystem, asReasoning, normalizeToolChoice, fitTools, fitted, renamed, unsaid, FILTER_SAYS, unflagged, usageOf, desktopHeld, explained, REFUSED_HINT, RESTRICTED_HINT, EXHAUSTED_HINT, REFUSED_DOC, commaPaused, commaPausedOn, opened, freeCredits, creditsOf, priceOf, fresh, SITES, renewing, renewed, SESSION, attend, scopeOf, turnKey, conversationKey, signatureOf, isMessages, withoutUnsigned, withoutSignedThinking, prepared, answeredBefore, nativeOf, modelOf, configModels, freeWindowOf, tokensShort, freeWindowRows, freeWindowLedger, readLedgerTail, pinned, rowTime, rowAccount, ofAccount, windowOffset, windowReset, FREE_WINDOW_MODELS, FREE_WINDOW_SPAN, FREE_WINDOW_GAP, FREE_WINDOW_LEAD, checkinOf, API_MESSAGES: NPM_MESSAGES, API_CHAT: NPM }
+export const _internal = { idFor, OFF_TAKEN, replyAfterToolsOn, withDefaultEffort, defaultEffortOf, gzipped, FREE_ONLY, PRICE_TTL, keyAccount, remoteConfig, cycleOf, typeUnions, systemAsUser, systemKey, DEFAULT_SYSTEM, EDGE_READS, fitShape, replyAfterTools, AFTER_TOOLS_REFUSED, enterpriseUsageOf, accountName, notDesktop, textOnly, runsOutOf, packagesOf, endOf, enterpriseHeaders, enterpriseOf, liveModels, browserSignIn, withSystem, asReasoning, normalizeToolChoice, fitTools, fitted, renamed, unsaid, FILTER_SAYS, unflagged, usageOf, desktopHeld, explained, REFUSED_HINT, RESTRICTED_HINT, EXHAUSTED_HINT, REFUSED_DOC, commaPaused, commaPausedOn, opened, freeCredits, creditsOf, priceOf, fresh, SITES, renewing, renewed, SESSION, attend, scopeOf, turnKey, conversationKey, signatureOf, isMessages, withoutUnsigned, withoutSignedThinking, prepared, answeredBefore, nativeOf, modelOf, configModels, freeWindowOf, tokensShort, freeWindowRows, freeWindowLedger, readLedgerTail, pinned, rowTime, rowAccount, ofAccount, windowOffset, windowReset, FREE_WINDOW_MODELS, FREE_WINDOW_SPAN, FREE_WINDOW_GAP, FREE_WINDOW_LEAD, checkinOf, API_MESSAGES: NPM_MESSAGES, API_CHAT: NPM }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@magpie-community/opencode-workbuddy-auth",
-  "version": "0.2.6",
+  "version": "0.2.8",
   "description": "OpenCode provider plugins for WorkBuddy's plans: WorkBuddy (CodeBuddy, copilot.tencent.com) and WorkBuddy AI (www.workbuddy.ai)",
   "type": "module",
   "main": "./index.mjs",