v0.10.5

What's Changed

Web Search Super‑Boost

We gave our web‑search agents a serious upgrade: every provider now returns clean, ready‑to‑use content and auto‑detects API keys from your environment. In short, searching the web feels like magic, and you won’t have to remember which secret goes where. #225 #229

Agent Workflow + Cost Transparency

Agents now whisper their cost and token usage back to you in every response, and the scheduler gracefully records failed runs on startup. No more “why did that run cost so much?” – your wallet gets an audit trail. #226 #215

CLI Polishing Party

The command‑line is now slicker than ever: bounded live panels for reasoning, adaptive buffering for tables/code, and a whole cascade of rendering polish (blank lines, delimiters, input flow tidying). It’s like the CLI got a fresh haircut and a new suit. #213 #211 #210 #209 #216 #218 #220

Chat Experience Revamp

Queued messages are now stacked line‑by‑line, you can type while the agent is busy, and the UI echoes “You:” when draining the queue. Chat feels more like a conversation and less like a monologue. #212 #220 #217 #221

LLM Provider Expansion

Added the coveted llama.cpp provider with configurable base URLs, plus client‑side reasoning‑tag parsing for local models. If you love running models on your own hardware, this is your moment. #199 #208

Config & Cost Tweaks

LLM retry counts are now configurable via jazz.config.json, and cost calculations are more precise. Fine‑tune reliability and budgeting without touching code. #228 #229 #224

Security Hardening

Eliminated a shell‑injection vector in the spawn‑argument handling. Your agents are safer from sneaky command‑line attacks. #227

CI & Docs Boost

CI now injects PR context into agents, fixes workspace paths, and prunes dead files. Documentation got a clarity sprint with plans for llama.cpp support. #207 #205 #219 #222 #223 #199 #191 #192

---

Commits

• 3b5f0c11 0.10.5 by @github-actions[bot]
• bb139601 feat(web-search): overhaul all providers to return content and follow latest best practices (#229) by @Landry Monga
• 34548bd0 fix(security): eliminate shell injection via spawn argument array (#227) by @Landry Monga
• 2ababe71 feat(config): make LLM retry count configurable via jazz.config.json (#228) by @Landry Monga
• 0824c3b3 ci(release): skip release if no commits since last tag by @lvndry
• 050e88ae chore(cost): more precise calculation by @lvndry
• 8a29a3ca feat(workflow): expose per-run cost and token usage in AgentResponse (#226) by @Landry Monga
• 62d00530 feat(web-search): auto-detect provider API keys from environment variables (#225) by @Landry Monga
• 01c645d7 fix(auto-update): replace changelog with releases link (#224) by @Landry Monga
• fa3286ba chore: improve vscode config (#223) by @Landry Monga
• d2505409 chore: remove TODO.md and README.proposed.md (#222) by @Landry Monga
• fa069c0f feat(history): session storage and conversation history (#221) by @Landry Monga
• ad57363f feat(chat): stack queued messages, one entry per line (#220) by @Landry Monga
• 9ba9840d feat(cli): bounded live panels for reasoning and subagents (#213) by @Landry Monga
• b674518f fix(chat): echo "You: " when draining the queued message (#217) by @Landry Monga
• fb4f804c chore: prune dead files and unused dependencies (#219) by @Landry Monga
• 17fdcec7 perf(cli): adaptive buffering for tables and code blocks (#218) by @Landry Monga
• d1150b41 feat(chat): queue messages typed while agent is busy (#212) by @Landry Monga
• 91972f78 fix(workflow): record failed scheduled runs and surface them on startup (#215) by @Landry Monga
• 861bcf20 perf(cli): buffer streaming deltas to ~80ms cadence (#216) by @Landry Monga
• 859929c5 fix(cli): rendering polish — blank line before metrics + reasoning delimiter (#211) by @Landry Monga
• e480f06b fix(cli): polish input flows — optional agent description + mask API key echo (#210) by @Landry Monga
• da1f4cc2 fix(cli): rebuild streaming pipeline on single-pending-buffer model (#209) by @Landry Monga
• 855c0af7 feat(llm): client-side reasoning-tag parsing for local models (#208) by @Landry Monga
• ac686d01 chore(docs): ignore AI agents docs by @lvndry
• bf4d2b26 docs(plan): implementation plan for llama.cpp provider support by @lvndry
• 7c0050f6 docs(spec): design for llama.cpp provider support by @lvndry
• 63a7bb34 feat(ci): give jazz agents PR context and finish the workspace fix (#207) by @Landry Monga
• c8fedc64 feat(llm): add llama.cpp provider and configurable base URLs (#199) by @Landry Monga
• 000c8b6c fix(ci): give pr‑assistant the workspace path and stop dumping raw failures (#205) by @Landry Monga
• 0d9ba9f6 fix(ci): use full namespaced model id for openrouter agents (#202) by @Landry Monga
• 229cc10b chore(release): OPENROUTER_API_KEY by @lvndry

Full diff

v0.10.4...v0.10.5