Repository navigation
Releases
v0.9.0
Safer by default. A cloned repo can no longer run commands when you open it: servers in a project's .ember/mcp.json start only after you run /mcp trust, and any change to that file needs trusting again. Print mode never trusts them. (#57) web_fetch now asks before fetching, because a URL can carry anything the agent has read. The prompt shows the full URL, and "always" covers one host. (#60) A command's whole process group is killed on a timeout or interrupt, so a hung cargo test or dev server no longer outlives it. A command that leaves a process running in the background no longer makes ember wait for the timeout. (#58)
Longer tasks. write_file writes several files in one call, all or nothing. When an edit's old_string matches more than one place, the error lists the line numbers. Five steps before the step limit, the agent is told to wrap up and summarize. ember -p -c "keep going" continues the last conversation in print mode, and --max-steps N (or EMBER_MAX_STEPS) sets the per-turn limit. (#55)
Math in the terminal. TeX in replies renders as Unicode: \(\mathrm{CH_4} + 2\mathrm{O_2} \rightarrow \mathrm{CO_2}\) shows as CH₄ + 2O₂ → CO₂, mhchem's \ce{} is supported, and a display \boxed{} gets a drawn box. Prices like $5 and $10 are left alone. (#50)
Pinned header. Once the banner scrolls away, a compact header with a mini mascot stays pinned at the top. It shows the mascot's mood, the model and the folder. Turn it off with EMBER_HEADER=off. (#52)
Other changes
/fixstops right away, with a clear message, when the project isn't committed in its git repository (for example, a folder inside an accidentalgit initin~). (#59)- The MIT copyright now names Guia Mae Bongulto too. (#56)
- Development:
cargo testno longer writes sessions into your real session history (#51), and releases are published from a single job (#61).
Full Changelog: v0.8.0...v0.9.0
v0.8.0
Fewer calls per task. edit_file takes several edits across files, applied all-or-nothing: a 3-file rename went from 14 calls and 65k input tokens to 6–7 calls and about 20k. (#32) The system prompt now carries a repo map of the project's definitions (#34), which now also covers C, C++, Java, Kotlin and C# and leaves out lockfiles and data files (#43), plus a rule to reproduce a bug before concluding there's nothing to fix (#22). On tinyledger, calls per task went from 7.6 to 5.3 and cost from $0.0044 to $0.0033.
Costs in dollars. The status bar shows the session's cost next to the token count, /usage has a cost column per model, and print mode's [usage] lines end with the cost. Cached input is priced at the cached rate. (#44) Commit, branch and PR drafting and /compact summaries use Lightning, about 5× cheaper. (#35)
Sturdier connections. Dropped connections, broken or stalled streams and 5xx errors are retried (#33), and replies that stop arriving fail after 180 s instead of hanging (#29). In print mode, a reply cut short by a retry is marked as interrupted instead of running into the retried one. (#45)
Look and feel
- New kawaii theme by default: pastel colors, sparkles, kaomoji and a mascot whose mood follows what ember is doing.
/theme classicbrings back the old look. Failed commands show ✗. (#26, #27) - The mascot is purple now. (#53)
- Small animations: the banner twinkles at startup, the mascot blinks at an idle prompt, the status line names the current step, a passing
/fixgets a burst of sparkles, and errors flash once. Turn them off with/animation off; they stay off whenCI,NO_COLORorEMBER_REDUCED_MOTION=1is set. (#54)
Other changes
/permissionslists what runs without asking, and/permissions savekeeps your "always" approvals for this project in~/.config/ember/permissions.json. (#48)- At the end of a turn, ember names any new files by their full path. (#47)
- File paths containing newlines or control characters are refused. (#49)
- The system prompt names the project's test command, and detection recognizes Django's
tests/runtests.pyand SymPy'sbin/test. (#19) EMBER_VERBOSE=1logs/fixbranch activity in print mode. (#45)- Development: end-to-end tests against a scripted model server (#30), a
/fixjudge accuracy eval (#42), four more tinyledger tasks (#46), and benchmark runner fixes (#23, #24, #25, #28).
New contributor: @gmvbongulto23 made their first contributions in #53 and #54.
Full Changelog: v0.7.0...v0.8.0
v0.7.0
Cheaper interactive sessions on Token Factory. Token Factory only credits cached tokens when everything up to the last user message matches an earlier request, so every follow-up used to re-bill the whole conversation. Follow-ups now reach the model as the result of an ask_user tool, which only extends the prompt. A replayed follow-up went from 0% to 98% cached. The model can also call ask_user itself to ask you a question. (#17)
More robust agent loop. Replies cut off at the token limit get a nudge to continue, calls to unknown tools get the list of valid tools, and broken tool arguments get a clear error instead of running with no arguments. (#18)
/fix
- Branches report back through a
run_teststool, so retry rounds stay cached. (#20) ember -p "/fix <issue>"runs the full pipeline non-interactively. Add--yesto apply the winning patch. (#20)
Other changes
/model ultrawarns that Ultra gets few cache hits on Token Factory: it caches in 4,352-token blocks and never counts the last one. (#15)EMBER_NO_WEB=1removes the web tools. (#16)- New SWE-bench Verified runner and tuning-loop tools in
bench/. The baseline with Super resolved 15 of 19 at $0.077 per resolved task. (#16)
Full Changelog: v0.6.0...v0.7.0
v0.6.0
What's Changed
- docs: restructure README, add badges, split config and tuning into docs/ by @aneesh-pradhan in #8
- chore: update .gitignore for the Rust rewrite by @aneesh-pradhan in #9
- Nemotron tuning: think budget, low_effort, sampling defaults; add bench by @aneesh-pradhan in #10
- ci: run clippy with warnings as errors by @aneesh-pradhan in #11
- Strip Nemotron's citation markers from replies by @aneesh-pradhan in #12
- Nemotron tuning round 2: keep reasoning by default, publish benchmark by @aneesh-pradhan in #13
- v0.6.0 by @aneesh-pradhan in #14
New Contributors
- @aneesh-pradhan made their first contribution in #8
Full Changelog: v0.5.0...v0.6.0
What's Changed
- docs: restructure README, add badges, split config and tuning into docs/ by @aneesh-pradhan in #8
- chore: update .gitignore for the Rust rewrite by @aneesh-pradhan in #9
- Nemotron tuning: think budget, low_effort, sampling defaults; add bench by @aneesh-pradhan in #10
- ci: run clippy with warnings as errors by @aneesh-pradhan in #11
- Strip Nemotron's citation markers from replies by @aneesh-pradhan in #12
- Nemotron tuning round 2: keep reasoning by default, publish benchmark by @aneesh-pradhan in #13
- v0.6.0 by @aneesh-pradhan in #14
New Contributors
- @aneesh-pradhan made their first contribution in #8
Full Changelog: v0.5.0...v0.6.0
What's Changed
- docs: restructure README, add badges, split config and tuning into docs/ by @aneesh-pradhan in #8
- chore: update .gitignore for the Rust rewrite by @aneesh-pradhan in #9
- Nemotron tuning: think budget, low_effort, sampling defaults; add bench by @aneesh-pradhan in #10
- ci: run clippy with warnings as errors by @aneesh-pradhan in #11
- Strip Nemotron's citation markers from replies by @aneesh-pradhan in #12
- Nemotron tuning round 2: keep reasoning by default, publish benchmark by @aneesh-pradhan in #13
- v0.6.0 by @aneesh-pradhan in #14
New Contributors
- @aneesh-pradhan made their first contribution in #8
Full Changelog: v0.5.0...v0.6.0
What's Changed
- docs: restructure README, add badges, split config and tuning into docs/ by @aneesh-pradhan in #8
- chore: update .gitignore for the Rust rewrite by @aneesh-pradhan in #9
- Nemotron tuning: think budget, low_effort, sampling defaults; add bench by @aneesh-pradhan in #10
- ci: run clippy with warnings as errors by @aneesh-pradhan in #11
- Strip Nemotron's citation markers from replies by @aneesh-pradhan in #12
- Nemotron tuning round 2: keep reasoning by default, publish benchmark by @aneesh-pradhan in #13
- v0.6.0 by @aneesh-pradhan in #14
New Contributors
- @aneesh-pradhan made their first contribution in #8
Full Changelog: v0.5.0...v0.6.0
v0.5.0
v0.4.0
Full Changelog: v0.3.0...v0.4.0
Full Changelog: v0.3.0...v0.4.0
Full Changelog: v0.3.0...v0.4.0
Full Changelog: v0.3.0...v0.4.0
v0.3.0
Full Changelog: v0.2.1...v0.3.0
Full Changelog: v0.2.1...v0.3.0
Full Changelog: v0.2.1...v0.3.0
Full Changelog: v0.2.1...v0.3.0
v0.2.1
Full Changelog: v0.2.0...v0.2.1
Full Changelog: v0.2.0...v0.2.1
Full Changelog: v0.2.0...v0.2.1
Full Changelog: v0.2.0...v0.2.1