chore: v1.2.2 — add staggered backoff for llama-server polling errors

README.md

@@ -53,6 +53,8 @@ The plugin discovers the llama-server URL by reading the OpenCode configuration
 
 Every 500ms, the plugin polls `GET /slots?model=<model>` on each discovered server. The model parameter is required by the `/slots` endpoint. If the model cannot be discovered from the current route's session, the plugin skips polling.
 
+When a server returns an error (e.g. HTTP 502) or is unreachable, the plugin backs off with an exponentially increasing delay per server (1s, 2s, 4s, ...) capped at 10s, with a small random jitter to stagger multiple servers. Backoff resets as soon as the server responds successfully again.
+
 ### State Classification
 
 Each processing slot is classified as prefill (PP) or generation (TG) by comparing `n_decoded` and `n_prompt_tokens` against per-slot/per-task baselines. The plugin tracks two baseline maps keyed by slot ID:
@@ -114,7 +116,7 @@ The following changes to the `/slots` endpoint would improve the plugin's functi
 
 For known issues, posting new ones, forking or contributing:
 
-https://codeberg.org/troed/oc-ls-stats
+https://git.sync.wtf/troed/oc-ls-stats
 
 ## Debug Logging
 

package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@troed/oc-ls-stats",
-  "version": "1.2.1",
+  "version": "1.2.2",
   "type": "module",
   "exports": {
     "./tui": {
@@ -9,7 +9,7 @@
     }
   },
   "scripts": {
-    "prepack": "npx -y tsx test-stats.ts"
+    "prepack": "npx -y tsx test-backoff.ts"
   },
   "engines": {
     "opencode": ">=1.3.14"

src/stats.ts

@@ -1,3 +1,7 @@
+export function backoffDelayMs(failures: number, baseMs: number = 1000, maxMs: number = 10000): number {
+  return Math.min(maxMs, baseMs * 2 ** (failures - 1))
+}
+
 export interface SlotState {
   id: number
   is_processing: boolean

test-backoff.ts

@@ -0,0 +1,28 @@
+import { test } from "node:test"
+import assert from "node:assert/strict"
+import { backoffDelayMs } from "./src/stats.ts"
+
+test("first failure retries after the base delay", () => {
+  assert.equal(backoffDelayMs(1), 1000)
+})
+
+test("second failure doubles the delay", () => {
+  assert.equal(backoffDelayMs(2), 2000)
+})
+
+test("delay grows exponentially", () => {
+  assert.equal(backoffDelayMs(3), 4000)
+  assert.equal(backoffDelayMs(4), 8000)
+})
+
+test("delay is capped at the maximum", () => {
+  assert.equal(backoffDelayMs(5), 10000)
+  assert.equal(backoffDelayMs(10), 10000)
+  assert.equal(backoffDelayMs(20), 10000)
+})
+
+test("uses provided base and max", () => {
+  assert.equal(backoffDelayMs(1, 200, 5000), 200)
+  assert.equal(backoffDelayMs(5, 200, 5000), 3200)
+  assert.equal(backoffDelayMs(20, 200, 5000), 5000)
+})

tui.tsx

@@ -2,7 +2,7 @@
 import type { TextRenderable } from "@opentui/core"
 import type { TuiPlugin, TuiPluginModule } from "@opencode-ai/plugin/tui"
 import { onCleanup } from "solid-js"
-import { processPoll } from "./src/stats.ts"
+import { backoffDelayMs, processPoll } from "./src/stats.ts"
 
 type StreamSample = {
   at: number
@@ -11,6 +11,8 @@ type StreamSample = {
 
 const STREAM_WINDOW_MS = 5_000
 const SLOT_POLL_MS = 500
+const BACKOFF_BASE_MS = 1_000
+const BACKOFF_MAX_MS = 10_000
 type MessageTiming = {
   sessionID: string
   requestStartAt: number
@@ -214,6 +216,7 @@ const tui: TuiPlugin = async (api) => {
 
   let llamaServerUrls: string[] = []
   let llamaServerModel: string | undefined
+  const backoffByUrl = new Map<string, { failures: number; nextRetryAt: number }>()
 
   const pollMetrics = async () => {
     if (!llamaServerUrls.length) return
@@ -230,8 +233,30 @@ const tui: TuiPlugin = async (api) => {
     }
 
     for (const baseUrl of llamaServerUrls) {
+      const now = Date.now()
+      const backoff = backoffByUrl.get(baseUrl)
+      if (backoff && now < backoff.nextRetryAt) {
+        tracker.failure = "n/a"
+        bump()
+        continue
+      }
+
       const slots = await fetchSlots(baseUrl, model)
-      const slotList = Array.isArray(slots) ? slots : slots ? Object.values(slots) : []
+      if (slots == null) {
+        const failures = (backoff?.failures ?? 0) + 1
+        const delay = backoffDelayMs(failures, BACKOFF_BASE_MS, BACKOFF_MAX_MS)
+        const jitter = Math.floor(Math.random() * (delay / 4))
+        backoffByUrl.set(baseUrl, {
+          failures,
+          nextRetryAt: Date.now() + Math.min(delay + jitter, BACKOFF_MAX_MS),
+        })
+        tracker.failure = "n/a"
+        bump()
+        continue
+      }
+
+      backoffByUrl.delete(baseUrl)
+      const slotList = Array.isArray(slots) ? slots : Object.values(slots)
       if (slotList.length === 0) {
         tracker.failure = "n/a"
         bump()