openclaw-code-agent 4.7.8 → 4.7.9

package/README.md

@@ -128,7 +128,7 @@ Because worktree isolation defaults to `delegate`, `defaultWorkdir` should norma
 
 Chat-launched sessions route updates back to their originating chat thread. For agent-launched tool sessions without an origin route, configure `fallbackChannel` or `agentChannels` in the reference guide.
 
-The current package targets and validates against OpenClaw `2026.7.1-2`. Its install, plugin API, Gateway, and peer dependency floor is `2026.7.1-2`, which includes both that numeric correction release and the higher-precedence stable `2026.7.1` release under SemVer ordering. The correction release only fixes singleton-array npm metadata handling for plugin installs and updates, so no host config migration or Code Agent runtime change is required. The manifest still declares `contracts.tools`, the plugin still imports only `openclaw/plugin-sdk/plugin-entry`, and its own session store, wake routing, callbacks, worktree flows, and Codex/Claude model restrictions remain plugin-owned. Current callback behavior around native callback data, routed follow-ups, retryable stale plan approval buttons, and serialized plan decisions remains unchanged.
+The current package targets and requires OpenClaw `2026.8.1`. The manifest still declares `contracts.tools`, the plugin still imports only `openclaw/plugin-sdk/plugin-entry`, and its session store, wake routing, callbacks, worktree flows, and Codex/Claude model restrictions remain plugin-owned. OpenClaw's host-side migration of `codex/*` and `openai-codex/*` model routes to `openai/*` does not rewrite Code Agent's unprefixed Codex harness models (`gpt-5.6-sol`, `gpt-5.6-terra`, and `gpt-5.6-luna`) or bypass `harnesses.codex.allowedModels`. Existing sessions, permissions, Telegram/topic routes, automation delivery context, callback ownership, and tool allowlists remain under the same plugin contracts; installing this package performs no host-side configuration migration.
 
 If you use Codex, make sure the local `codex` command or `OPENCLAW_CODEX_APP_SERVER_COMMAND` override is available and authenticated. Codex-specific defaults live under `harnesses.codex`: `reasoningEffort` is sent as `reasoningEffort`, and `fastMode: true` sends `service_tier: "fast"` on Codex App Server thread, resume, and turn payloads. The Codex harness starts the app server with stdio listener args by default, only sends UUID-shaped backend thread IDs to `thread/resume`, and reports startup timeouts with redacted recent stderr. When Codex auth is inconsistent, this is the recommended `~/.codex/config.toml` setting:
 
@@ -136,6 +136,8 @@ If you use Codex, make sure the local `codex` command or `OPENCLAW_CODEX_APP_SER
 forced_login_method = "chatgpt"
 ```
 
+For Codex sessions authenticated with an OpenAI API key, the plugin estimates token charges from the App Server's per-response usage and shows the cumulative amount in session listings and `agent_stats`. ChatGPT OAuth/subscription sessions remain `$0` because they are quota-backed rather than pay-as-you-go API calls. The estimator covers the built-in GPT-5.6 Sol, Terra, and Luna models, including cached input, cache writes, Fast mode, and long-context rates; unknown models remain unpriced. Reasoning tokens are already included in App Server `outputTokens` and are not added a second time.
+
 If you use experimental OpenCode, make sure local `opencode >= 1.16.2` is available and configured with provider auth. The plugin starts `opencode serve` per session on localhost and uses OpenCode's classic session lifecycle routes while v2 session wait remains unavailable. Leave `harnesses.opencode.defaultModel` unset to let OpenCode choose its configured provider default, or pass an explicit `provider/model` string for a launch.
 
 ## First Session