@elvatis_com/openclaw-cli-bridge-elvatis 1.8.4 → 1.8.6

package/README.md

@@ -2,7 +2,7 @@
 
 > OpenClaw plugin that bridges locally installed AI CLIs (Codex, Gemini, Claude Code) as model providers — with slash commands for instant model switching, restore, health testing, and model listing.
 
-**Current version:** `1.8.2`
+**Current version:** `1.8.6`
 
 ---
 
@@ -32,36 +32,36 @@ All commands use gateway-level `commands.allowFrom` for authorization (`requireA
 
 **Claude Code CLI** (routed via local proxy on `:31337`):
 
-| Command | Model |
-|---|---|
-| `/cli-sonnet` | `vllm/cli-claude/claude-sonnet-4-6` |
-| `/cli-opus` | `vllm/cli-claude/claude-opus-4-6` |
-| `/cli-haiku` | `vllm/cli-claude/claude-haiku-4-5` |
+| Command | Model | Notes |
+|---|---|---|
+| `/cli-sonnet` | `vllm/cli-claude/claude-sonnet-4-6` | ✅ Tested |
+| `/cli-opus` | `vllm/cli-claude/claude-opus-4-6` | ✅ Tested |
+| `/cli-haiku` | `vllm/cli-claude/claude-haiku-4-5` | ✅ Tested |
 
 **Gemini CLI** (routed via local proxy on `:31337`, stdin + `cwd=/tmp`):
 
-| Command | Model |
-|---|---|
-| `/cli-gemini` | `vllm/cli-gemini/gemini-2.5-pro` |
-| `/cli-gemini-flash` | `vllm/cli-gemini/gemini-2.5-flash` |
-| `/cli-gemini3` | `vllm/cli-gemini/gemini-3-pro-preview` |
-| `/cli-gemini3-flash` | `vllm/cli-gemini/gemini-3-flash-preview` |
+| Command | Model | Notes |
+|---|---|---|
+| `/cli-gemini` | `vllm/cli-gemini/gemini-2.5-pro` | ✅ Tested |
+| `/cli-gemini-flash` | `vllm/cli-gemini/gemini-2.5-flash` | ✅ Tested |
+| `/cli-gemini3` | `vllm/cli-gemini/gemini-3-pro-preview` | ✅ Tested |
+| `/cli-gemini3-flash` | `vllm/cli-gemini/gemini-3-flash-preview` | ✅ Tested |
 
 **Codex CLI** (via `openai-codex` provider — OAuth auth, calls OpenAI API directly, **not** through the local proxy):
 
 | Command | Model | Notes |
 |---|---|---|
 | `/cli-codex` | `openai-codex/gpt-5.3-codex` | ✅ Tested |
-| `/cli-codex-spark` | `openai-codex/gpt-5.3-codex-spark` | |
-| `/cli-codex52` | `openai-codex/gpt-5.2-codex` | |
+| `/cli-codex-spark` | `openai-codex/gpt-5.3-codex-spark` | ✅ Tested |
+| `/cli-codex52` | `openai-codex/gpt-5.2-codex` | ✅ Tested |
 | `/cli-codex54` | `openai-codex/gpt-5.4` | May require upgraded OAuth scope |
 | `/cli-codex-mini` | `openai-codex/gpt-5.1-codex-mini` | ✅ Tested |
 
 **BitNet local inference** (via local proxy → llama-server on 127.0.0.1:8082, no API key):
 
-| Command | Model |
-|---|---|
-| `/cli-bitnet` | `vllm/local-bitnet/bitnet-2b` |
+| Command | Model | Notes |
+|---|---|---|
+| `/cli-bitnet` | `vllm/local-bitnet/bitnet-2b` | ✅ Tested |
 
 **Utility:**
 
@@ -91,14 +91,16 @@ All commands use gateway-level `commands.allowFrom` for authorization (`requireA
 
 Routes requests through real browser sessions on the provider's web UI. Requires a valid login session (free or paid tier). Uses persistent Chromium profiles — sessions survive gateway restarts.
 
+> **Note:** Only Grok and Gemini are active browser providers. Claude and ChatGPT browser routes were removed in v1.6.x — use `cli-claude/*` (Claude CLI) and `openai-codex/*` / `copilot-proxy` instead.
+
 **Grok** (grok.com — SuperGrok subscription):
 
 | Model | Notes |
 |---|---|
-| `web-grok/grok-3` | Full model |
-| `web-grok/grok-3-fast` | Faster variant |
-| `web-grok/grok-3-mini` | Lightweight |
-| `web-grok/grok-3-mini-fast` | Fastest |
+| `web-grok/grok-3` | Full model | ✅ Tested |
+| `web-grok/grok-3-fast` | Faster variant | ✅ Tested |
+| `web-grok/grok-3-mini` | Lightweight | ✅ Tested |
+| `web-grok/grok-3-mini-fast` | Fastest | ✅ Tested |
 
 | Command | What it does |
 |---|---|
@@ -110,10 +112,10 @@ Routes requests through real browser sessions on the provider's web UI. Requires
 
 | Model | Notes |
 |---|---|
-| `web-gemini/gemini-2-5-pro` | Gemini 2.5 Pro |
-| `web-gemini/gemini-2-5-flash` | Gemini 2.5 Flash |
-| `web-gemini/gemini-3-pro` | Gemini 3 Pro |
-| `web-gemini/gemini-3-flash` | Gemini 3 Flash |
+| `web-gemini/gemini-2-5-pro` | Gemini 2.5 Pro | ✅ Tested |
+| `web-gemini/gemini-2-5-flash` | Gemini 2.5 Flash | ✅ Tested |
+| `web-gemini/gemini-3-pro` | Gemini 3 Pro | ✅ Tested |
+| `web-gemini/gemini-3-flash` | Gemini 3 Flash | ✅ Tested |
 
 | Command | What it does |
 |---|---|
@@ -121,37 +123,9 @@ Routes requests through real browser sessions on the provider's web UI. Requires
 | `/gemini-status` | Show session validity + cookie expiry |
 | `/gemini-logout` | Clear session |
 
-**Claude.ai** (claude.ai — Pro/Team subscription):
+**Claude.ai** ~~(removed in v1.6.x)~~ — use `/cli-sonnet`, `/cli-opus`, `/cli-haiku` instead.
 
-| Model | Notes |
-|---|---|
-| `web-claude/claude-sonnet` | Claude Sonnet (web) |
-| `web-claude/claude-opus` | Claude Opus (web) |
-| `web-claude/claude-haiku` | Claude Haiku (web) |
-
-| Command | What it does |
-|---|---|
-| `/claude-login` | Authenticate, save cookies to `~/.openclaw/claude-profile/` |
-| `/claude-status` | Show session validity + cookie expiry |
-| `/claude-logout` | Clear session |
-
-**ChatGPT** (chatgpt.com — Plus/Pro subscription):
-
-| Model | Notes |
-|---|---|
-| `web-chatgpt/gpt-4o` | GPT-4o |
-| `web-chatgpt/gpt-4o-mini` | GPT-4o Mini |
-| `web-chatgpt/gpt-4.1` | GPT-4.1 |
-| `web-chatgpt/o3` | o3 |
-| `web-chatgpt/o4-mini` | o4-mini |
-| `web-chatgpt/gpt-5` | GPT-5 |
-| `web-chatgpt/gpt-5-mini` | GPT-5 Mini |
-
-| Command | What it does |
-|---|---|
-| `/chatgpt-login` | Authenticate, save cookies to `~/.openclaw/chatgpt-profile/` |
-| `/chatgpt-status` | Show session validity + cookie expiry |
-| `/chatgpt-logout` | Clear session |
+**ChatGPT** ~~(removed in v1.6.x)~~ — use `/cli-codex`, `openai-codex/*`, or `copilot-proxy` instead.
 
 **Session lifecycle:**
 - First use: run `/xxx-login` once — authenticates and saves cookies to persistent Chromium profile
@@ -368,6 +342,19 @@ npm test            # vitest run (83 tests)
 
 ## Changelog
 
+### v1.8.6
+- **fix:** Remove stale web-claude/* and web-chatgpt/* entries from model list (status page showed removed providers)
+
+/
+### v1.8.5
+- **fix:** Replace full system prompt with 30-token mini stub for BitNet (prevents context overflow)
+- **fix:** Truncate to last 6 non-system messages before forwarding to BitNet (4096 token limit)
+- **fix:** Flatten multi-part content arrays to plain strings (llama-server crash fix)
+
+### v1.8.3 → v1.8.4
+- Intermediate BitNet crash fixes (superseded by v1.8.5)
+
+/
 ### v1.8.2
 - **fix:** `local-bitnet/*` exempt from tool-call rejection — llama-server ignores tool schemas silently. OpenClaw always sends tools with every request, so this was blocking all BitNet usage.
 

package/SKILL.md

@@ -68,4 +68,4 @@ On gateway restart, if any session has expired, a **WhatsApp alert** is sent aut
 
 See `README.md` for full configuration reference and architecture diagram.
 
-**Version:** 1.8.2
+**Version:** 1.8.6

package/openclaw.plugin.json

@@ -1,7 +1,8 @@
 {
   "id": "openclaw-cli-bridge-elvatis",
+  "slug": "openclaw-cli-bridge-elvatis",
   "name": "OpenClaw CLI Bridge",
-  "version": "1.8.2",
+  "version": "1.8.6",
   "license": "MIT",
   "description": "Phase 1: openai-codex auth bridge. Phase 2: local HTTP proxy routing model calls through gemini/claude CLIs (vllm provider).",
   "providers": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@elvatis_com/openclaw-cli-bridge-elvatis",
-  "version": "1.8.4",
+  "version": "1.8.6",
   "description": "Bridges gemini, claude, and codex CLI tools as OpenClaw model providers. Reads existing CLI auth without re-login.",
   "type": "module",
   "openclaw": {

package/src/proxy-server.ts

@@ -94,18 +94,8 @@ export const CLI_MODELS = [
   { id: "web-gemini/gemini-2-5-flash", name: "Gemini 2.5 Flash (web session)", contextWindow: 1_000_000, maxTokens: 8192 },
   { id: "web-gemini/gemini-3-pro",     name: "Gemini 3 Pro (web session)",     contextWindow: 1_000_000, maxTokens: 8192 },
   { id: "web-gemini/gemini-3-flash",   name: "Gemini 3 Flash (web session)",   contextWindow: 1_000_000, maxTokens: 8192 },
-  // Claude web-session models (requires /claude-login)
-  { id: "web-claude/claude-sonnet",     name: "Claude Sonnet (web session)",     contextWindow: 200_000, maxTokens: 8192 },
-  { id: "web-claude/claude-opus",       name: "Claude Opus (web session)",       contextWindow: 200_000, maxTokens: 8192 },
-  { id: "web-claude/claude-haiku",      name: "Claude Haiku (web session)",      contextWindow: 200_000, maxTokens: 8192 },
-  // ChatGPT web-session models (requires /chatgpt-login)
-  { id: "web-chatgpt/gpt-4o",           name: "GPT-4o (web session)",            contextWindow: 128_000, maxTokens: 16_384 },
-  { id: "web-chatgpt/gpt-4o-mini",      name: "GPT-4o Mini (web session)",       contextWindow: 128_000, maxTokens: 16_384 },
-  { id: "web-chatgpt/gpt-4.1",          name: "GPT-4.1 (web session)",           contextWindow: 1_047_576, maxTokens: 32_768 },
-  { id: "web-chatgpt/o3",               name: "o3 (web session)",                contextWindow: 200_000, maxTokens: 100_000 },
-  { id: "web-chatgpt/o4-mini",          name: "o4-mini (web session)",           contextWindow: 200_000, maxTokens: 100_000 },
-  { id: "web-chatgpt/gpt-5",            name: "GPT-5 (web session)",             contextWindow: 1_047_576, maxTokens: 32_768 },
-  { id: "web-chatgpt/gpt-5-mini",       name: "GPT-5 Mini (web session)",        contextWindow: 1_047_576, maxTokens: 32_768 },
+  // Claude → use cli-claude/* instead (web-claude removed in v1.6.x)
+  // ChatGPT → use openai-codex/* or copilot-proxy instead (web-chatgpt removed in v1.6.x)
   // ── Local BitNet inference ──────────────────────────────────────────────────
   { id: "local-bitnet/bitnet-2b",       name: "BitNet b1.58 2B (local CPU inference)", contextWindow: 4_096, maxTokens: 2_048 },
 ];
@@ -578,15 +568,17 @@ async function handleRequest(
       // BitNet has a 4096 token context window. Long sessions blow it up and
       // cause a hard C++ crash (no graceful error). Truncate to system prompt +
       // last 10 messages (~2k tokens max) to stay safely within the limit.
-      const BITNET_MAX_MESSAGES = 10;
+      const BITNET_MAX_MESSAGES = 6;
+      // Replace the full system prompt (MEMORY.md etc, ~2k+ tokens) with a
+      // minimal one so BitNet's 4096-token context isn't blown by the system msg alone.
+      const BITNET_SYSTEM = "You are Akido, a concise AI assistant. Answer briefly and directly. Current user: Emre. Timezone: Europe/Berlin.";
       const allFlat = parsed.messages.map((m) => ({
         role: m.role,
         content: flattenContent(m.content),
       }));
-      const systemMsgs = allFlat.filter((m) => m.role === "system");
       const nonSystemMsgs = allFlat.filter((m) => m.role !== "system");
       const truncated = nonSystemMsgs.slice(-BITNET_MAX_MESSAGES);
-      const bitnetMessages = [...systemMsgs, ...truncated];
+      const bitnetMessages = [{ role: "system", content: BITNET_SYSTEM }, ...truncated];
       const requestBody = JSON.stringify({ ...parsed, messages: bitnetMessages, tools: undefined });
 
       try {