Release v0.17.2

Changes from 0.17.1 to 0.17.2

What's Changed

• feat(headless): exclude AskUserQuestion from headless mode toolset by @kl2806 in #1266
• fix(model): align free-tier model dropdown behavior and default to GLM-5 by @sarahwooders in #1263
• fix(tui): don't clear conversation model override on transient retrieve errors by @jnjpng in #1269
• fix(ws): remote interrupt recovery for listen-client [LET-7796] by @cpacker in #1272
• fix(tui): link footer subagent by @christinatong01 in #1270
• feat(listen): emit run_request_error when initial POST fails by @4shub in #1267
• feat: add GPT-5.4 model support by @kl2806 in #1275
• fix: context window gpt 5.4 by @AriWebb in #1276
• nit: beta for auto models by @AriWebb in #1278
• chore: remove system-append flag by @devanshrj in #1280
• fix(tui): better footer subagent link and alignment by @christinatong01 in #1283
• fix(tui): keep conversation model overrides sticky by @paulbettner in #1238
• chore: restore null! placeholders for forward-reference pattern by @jnjpng in #1284
• fix: send max_tokens null to explicitly unset stale values on model switch by @jnjpng in #1281
• fix(listen): recover stale approval conflicts after stop-reason errors by @cpacker in #1285
• feat(tui): style bg subagent footer by @christinatong01 in #1288
• fix(listen): normalize interrupt tool returns for websocket parity by @cpacker in #1289
• revert: "fix(listen): normalize interrupt tool returns for websocket parity" by @cpacker in #1291
• fix: restrict --system flag to known preset IDs by @devanshrj in #1290
• fix: use SYSTEM_PROMPT for default preset instead of letta-claude by @devanshrj in #1292
• fix(listen): preserve interrupt error status through next-turn persistence by @cpacker in #1294
• fix(listen): expose interrupted tool state for reconnect hydration by @cpacker in #1295
• feat(message): add OpenAI websocket responses mode header by @jnjpng in #1297
• fix: don't pass null max_output_tokens in model_settings by @jnjpng in #1299
• feat: add gpt-5.4-fast by @AriWebb in #1300
• feat(model): add GPT-5.4 and GPT-5.4 Pro to model selector by @sarahwooders in #1301
• feat: send image version in telemetry by @carenthomas in #1298
• fix(tui): gate debug footer stub behind LETTA_DEBUG_FOOTER env var by @cpacker in #1302
• chore: resync version to 0.17.1 to match npm registry by @cpacker in #1303

Full Changelog: v0.17.0...v0.17.2