v0.3.352

What's Changed

• Add cheapest launch profile and --search opt-in for semantic search by @animeshkundu in #158
• Enforce router-only surfaces on pinned profiles (fast/cheap/cheap1m/cheapest) by @animeshkundu in #159
• Pin cheap-family subagents to 200K with non-catalog alias identities by @animeshkundu in #160
• Fix shim non-streaming AIC double-count by @animeshkundu in #161
• Group tool-result images after the tool block on the Anthropic shim by @animeshkundu in #162
• Fix AIC streaming undercount: priced taps skip interim zero-nano frames by @animeshkundu in #163
• Re-emit priced copilot_usage on terminal stream deltas by @animeshkundu in #164
• Tune launch modes: add balanced profile, drop implementer, rename General-Purpose by @animeshkundu in #165
• Add gh-gather-context, gh-plan, gh-implement pipeline skills for pinned profiles by @animeshkundu in #166
• Rewrite SWE pipeline skills to native-only subagents by @animeshkundu in #167
• Code search next arch: lexical fixes, C# grammar, rerank, serve-while-stale, index CLI, service track by @animeshkundu in #168
• Semantic service: static OpenBLAS + CUDA variants with CPU fallback by @animeshkundu in #169
• Generate search prompts from the --search flag: semantic-first when enabled, lexical-only otherwise by @animeshkundu in #170
• index: service-default backend with delta indexing and full-compute foreground builds by @animeshkundu in #171
• Add --bluebird flag routing semantic + lexical code search through Bluebird MCP by @animeshkundu in #174
• Swap balanced General-Purpose and reviewer models with efforts by @animeshkundu in #175
• Make balanced profile advisor-free by @animeshkundu in #176
• Fix Advisor context usage and AIC accounting by @animeshkundu in #177
• Bump default models to gpt-6-sol, gpt-6-luna, and claude-opus-5.5 by @animeshkundu in #178
• Switch cheapest and balanced reviewer to gpt-6-sol high at 200K by @animeshkundu in #179
• Remove Plan subagent from cheapest and balanced; cheapest lead reviews final plan with Advisor by @animeshkundu in #180
• Run worker-browse on gpt-6-luna at low effort in the 200K default window by @animeshkundu in #181
• Add github-router pi launcher for cheapest/balanced Pi CLI modes by @animeshkundu in #182
• Drive Pi footer from the shared internal-aic-status runner for Claude-exact statusline parity by @animeshkundu in #183
• Bridge Copilot + Claude memory into Pi via mirror AGENTS.md by @animeshkundu in #184
• Fix Pi bridge: workspace fallback, oracle query mapping, stripped browser names, NUL sanitization by @animeshkundu in #185
• Advertise Pi vision + helpful bundle at lowest cost by @animeshkundu in #186
• Account Pi AIC totals holistically with Claude-parity exit summary by @animeshkundu in #187
• Drop pi-agent-extensions files picker (ctrl+shift+o clash) and cache Pi npm packages across launches by @animeshkundu in #188
• Default to gpt-5.6-sol with gpt-6-sol as fallback by @animeshkundu in #189
• Cheapest Luna/max reviewer with capped Sol protege Advisor, Sol prompt rewrite, and tier-tuned prompts by @animeshkundu in #190
• Show the context window beside the bar in the shared status line by @animeshkundu in #191
• Add user-only /gh-rebrief on-demand cheapest re-brief with 4K extractive context by @animeshkundu in #192
• Pin cheap/cheapest/balanced to strict 200K via CLAUDE_CODE_DISABLE_1M_CONTEXT by @animeshkundu in #193
• Add ADO and Kusto MCP groups, Pi 1.3.0 parity, and dependency modernization by @animeshkundu in #194

Full Changelog: v0.3.315...v0.3.352