Skip to content

v0.10.8

Choose a tag to compare

View all tags
@lvndry lvndry released this 03 May 12:38
· 27 commits to main since this release
v0.10.8
f8a3256

What's Changed

Tool Aliases & Prompt Boosts

We’ve given the CLI a swagger makeover: tools now understand aliases, so you can shout “jazz tmux” instead of typing the full command name. Plus, prompts are now smarter about blocked interpreter flags – your agents won’t get stuck on a stubborn flag ever again. 🎉

Web Search Superpowers

The web‑search engine got a full‑blown upgrade! Native OpenRouter support means the agent can fetch and search the web straight from the server. All providers now return content, not just URLs, and they automatically sniff out API keys from your environment. Searching is now as easy as “Ask Jazz to google it” – and it actually brings back the juicy bits. 🌐

LLM & Agent Power‑Ups

  • Raised the LLM request timeout floor to 10 minutes, so those marathon reasoning sessions never time out.
  • Bumped default retry attempts from 3 to 8, and made the retry count configurable via jazz.config.json.
  • Added a brand‑new llama.cpp provider with configurable base URLs – you can now run local LLMs without leaving your terminal.
  • Client‑side reasoning‑tag parsing makes local models smarter about “thinking aloud”.
  • Security hardened the spawn function to eliminate shell‑injection vectors.

All together, your agents are now more resilient, configurable, and safer.

Workflow & Cost Visibility

Workflows now brag about their per‑run cost and token usage in the AgentResponse, letting you keep an eye on budget while automating. Failed scheduled runs are recorded and shown at startup, so nothing slips through the cracks. 🚀

CLI Polish & Performance

The CLI got a performance spa: adaptive buffering for tables/code blocks, ~80 ms streaming delta cadence, and polished rendering (blank lines before metrics, clearer reasoning delimiters). Live panels are now bounded, giving you a smoother, less jittery interactive experience.

CI/CD Integrations

  • PR triggers (/jazz and /jazz-review) are now built‑in, and API keys are automatically redacted from terminal history.
  • The release pipeline can now skip releases when there are no new commits, keeping the tag history tidy.
  • CI now passes workspace paths to the PR assistant and stops dumping raw failures, making debugging CI runs painless.

Docs & Community

A fresh README revamp shines a spotlight on the new features, and we added detailed design plans for the upcoming llama.cpp provider. Your onboarding experience just got a whole lot brighter.

Behind‑the‑Scenes Upkeep

All dependencies have been refreshed to the latest versions, and the cost calculator now offers more precise pricing. Dead files and unused deps have been pruned, and the VS Code config got a tidy‑up.

Commits

  • f8a3256 0.10.8
  • 9f1091f feat(tools): add tool alias support + prompt improvements for blocked interpreter flags (#233)
  • b4c5a5b docs: README (#234)
  • 6631b4e chore(deps): update all dependencies (#232)
  • e7a1565 fix(agent): raise LLM request timeout floor to 10 min
  • c0eacf8 fix(agent): bump DEFAULT_MAX_LLM_RETRIES from 3 to 8 (#231)
  • 2d4a909 feat(openrouter): native web_search and web_fetch server tool support (#230)
  • bb13960 feat(web-search): overhaul all providers to return content and follow latest best practices (#229)
  • 34548bd fix(security): eliminate shell injection via spawn argument array (#227)
  • 2ababe7 feat(config): make LLM retry count configurable via jazz.config.json (#228)
  • 0824c3b ci(release): skip release if no commits since last tag
  • 050e88a chore(cost): more precise calculation
  • 8a29a3c feat(workflow): expose per-run cost and token usage in AgentResponse (#226)
  • 62d0053 feat(web-search): auto-detect provider API keys from environment variables (#225)
  • 01c645d fix(auto-update): replace changelog with releases link (#224)
  • fa3286b chore: improve vscode config (#223)
  • d250540 chore: remove TODO.md and README.proposed.md (#222)
  • fa069c0 feat(history): session storage and conversation history (#221)
  • ad57363 feat(chat): stack queued messages, one entry per line (#220)
  • 9ba9840 feat(cli): bounded live panels for reasoning and subagents (#213)
  • b674518 fix(chat): echo "You: " when draining the queued message (#217)
  • fb4f804 chore: prune dead files and unused dependencies (#219)
  • 17fdcec perf(cli): adaptive buffering for tables and code blocks (#218)
  • d1150b4 feat(chat): queue messages typed while agent is busy (#212)
  • 91972f7 fix(workflow): record failed scheduled runs and surface them on startup (#215)
  • 861bcf2 perf(cli): buffer streaming deltas to ~80ms cadence (#216)
  • 859929c fix(cli): rendering polish — blank line before metrics + reasoning delimiter (#211)
  • e480f06 fix(cli): polish input flows — optional agent description + mask API key echo (#210)
  • da1f4cc fix(cli): rebuild streaming pipeline on single-pending-buffer model (#209)
  • 855c0af feat(llm): client-side reasoning-tag parsing for local models (#208)
  • ac686d0 docs(plan): implementation plan for llama.cpp provider support
  • bf4d2b2 docs(plan): design for llama.cpp provider support
  • 63a7bb3 feat(ci): give jazz agents PR context and finish the workspace fix (#207)
  • c8fedc6 feat(llm): add llama.cpp provider and configurable base URLs (#199)
  • 000c8b6 fix(ci): give pr‑assistant the workspace path and stop dumping raw failures (#205)
  • 0d9ba9f fix(ci): use full namespaced model id for openrouter agents (#202)
  • 229cc10 chore(release): OPENROUTER_API_KEY
  • f20e237 Audit follow‑ups: persona profiles, security, tokenizer, skills bloat, markdown (#198)
  • 3b9d4a5 feat(ci,cli): /jazz PR trigger; redact API keys in terminal history (#196)
  • 3fe3735 feat(cli,ci): add @Jazz PR assistant and polish output (#195)

Full diff

v0.10.7...v0.10.8

Contributors

Jazz
Assets 2
Loading