v0.1.7

• Discontinued OpenClaw Quick Replies after OpenClaw 2026.8.1 introduced first-party structured questions with native controls, validated Gateway answers, and broader channel support.
• Replaced the active installation and configuration guide with an archival notice, an honest comparison of the remaining legacy behavior, and migration instructions for native ask_user questions.
• Updated package, catalog, support, and security metadata to make the end of maintenance visible. Runtime behavior is unchanged from v0.1.6.

v0.1.6

• Renamed the user-facing OpenClaw and ClawHub plugin title from "OpenClaw Quick Replies" to "Quick Replies" while preserving stable package and plugin identifiers.
• Explained expired or stale update and restart approvals in Telegram, preserved the restart control after duplicate install callbacks, and deduplicated concurrent Gateway restart requests.
• Updated the OpenClaw compatibility floor and development SDK to 2026.7.1-2, including its corrected npm plugin metadata handling.
• Updated CI and transitive dependency resolutions to current patched releases.

v0.1.5

• Fixed update-result delivery so a failed Telegram message edit falls back to a normal reply, preserving the Gateway restart control after a successful installation.
• Updated customer-facing references to consistently use the full OpenClaw Quick Replies name.

v0.1.4

• Added a supported thinkLevel evaluator setting, defaulting to the lowest nonzero minimal level and passed directly to OpenClaw's embedded model run, with cache isolation, documentation, and regression coverage.
• Validated evaluator thinking levels against the selected model's OpenClaw runtime policy and fail open instead of silently substituting an unsupported level.
• Confirmed explicit numbered and bulleted choices continue through the model evaluator rather than a deterministic fast path, and added a credential-free plugin-overhead benchmark.

v0.1.3

• Fixed token-backed ClawHub publishing to use the CLI's documented temporary configuration file while keeping the repository secret out of artifacts and logs.
• Fixed self-updates to force-reinstall the approved exact version from the existing npm or ClawHub source, verify the resulting installed version, and log privacy-safe Telegram callback outcomes before offering a separately authorized Gateway restart.
• Reduced Quick Replies evaluator overhead with raw model-run mode, timeout cancellation, semantic cross-message caching, an MCP-free per-run OpenClaw config projection, and bounded stage timing diagnostics.
• Documented the OpenClaw 2026.7.1 Codex lifecycle defect that still inherits user-level Codex MCP servers for tool-disabled temporary threads and requires a core fix for complete resource cleanup.

v0.1.2

• Added a daily, non-blocking update check that offers authorized Telegram controls for installing the exact approved version through OpenClaw's native plugin manager, preserving both npm and ClawHub install sources, and separately confirming a Gateway restart.
• Added an updateChecks setting, enabled by default, and documented the npm registry metadata request used by the checker.
• Fixed release automation to publish the exact verified tarball to npm and ClawHub, retain verified artifacts through delayed environment approval, and attribute the pre-tag ClawHub dry run to the exact commit without claiming a not-yet-created release ref.
• Clarified how operator-triggered and centrally scheduled native plugin updates work.
• Made Gateway restart confirmations single-use with a one-hour expiry, and suppressed update notices whenever Telegram channel data already contains interactive controls.

v0.1.1

• Increased the default evaluator timeout from five seconds to 20 seconds so managed evaluators have enough time to add quick reply buttons before the original message is sent without them.

v0.1.0

• Added conservative, model-evaluated Telegram quick reply suggestions.
• Added strict canonical oqr:v1:<base64url> callbacks with a 42-byte UTF-8 value limit.
• Added authorization, source-context, duplicate-selection, and button-cleanup safeguards.
• Added explicit-ask filtering, a configurable five-second timeout, concurrent request deduplication, and a bounded five-minute decision cache.
• Added public ClawHub/npm packaging, release validation, documentation, screenshots, and security guidance.
• Documented Discord as unsupported until OpenClaw provides a generic inbound-submission callback contract.