Skip to content

v0.0.1

Latest

Choose a tag to compare

@github-actions github-actions released this 18 Nov 22:08
532c215

What's Changed

  • Scaffold token-max with quest, a Nix dev shell, and CI by @kixelated in #1
  • Plan TypeScript CLI, cost checks, and npm distribution by @kixelated in #2
  • Configure protected squash auto-merge by @kixelated in #4
  • Implement local TypeScript replay and Claude TTL recommendations by @kixelated in #3
  • Measure shared component exposure and invocation evidence by @kixelated in #5
  • feat!: Recommend compaction windows net of summary and prefix costs by @kixelated in #6
  • Polished terminal report by @kixelated in #7
  • Compare native idle compaction on/off net costs by @kixelated in #8
  • Prepare one-command npm distribution and protected release path by @kixelated in #9
  • Plan T3 settings and explicit SSH audits by @kixelated in #11
  • Recommend standard speed and report API-equivalent speed premiums by @kixelated in #10
  • Give a verdict per setting from real histories by @kixelated in #19
  • Price every piece of recurring context by @kixelated in #20
  • One line per setting, with cost-per-use recommendations by @kixelated in #21
  • Rewrite token-max in Rust by @kixelated in #22
  • Flag a foreground shell timeout cap past the cache TTL by @kixelated in #25
  • Price Codex context and tool output by @kixelated in #26
  • Audit SSH hosts with their installed token-max by @kixelated in #29
  • Estimate background generation costs per feature by @kixelated in #23
  • Plan conditional Codex pricing for missing service tiers by @kixelated in #30
  • Plan project-scoped background controls by @kixelated in #31
  • Switch each Claude skill with the setting its source supports by @kixelated in #28
  • Price capped keepalive pings against idle sessions and idle compaction by @kixelated in #24
  • feat!: compact JSON and shared evidence for large histories by @kixelated in #27
  • Estimate Codex API costs with explicit tier assumptions by @kixelated in #34
  • Add T3 Claude and Codex profiles, controls and idle-policy evidence by @kixelated in #36
  • Report MCP tool costs, loading controls and instruction budgets by @kixelated in #33
  • fix!: resolve background controls per recorded project by @kixelated in #32
  • Report Codex per-skill catalog costs, usage and duplicate controls by @kixelated in #35
  • fix: make the report complete on real multi-harness histories by @kixelated in #37
  • chore(quest): plan restoring the large-history runtime by @kixelated in #38
  • chore(quest): plan T3 state reading and text-generation model audits by @kixelated in #39
  • fix: read real T3 state through its indexes and name the failure cause by @kixelated in #41
  • feat: report T3's text-generation model by @kixelated in #42
  • chore(quest): plan the ready-for-review Check race by @kixelated in #43
  • perf: stop cloning profiles per idle scope by @kixelated in #40
  • fix(ci): read draft state when the Check run starts by @kixelated in #45
  • feat: give subagents their root's T3 owner and name unresolved causes by @kixelated in #44
  • chore(quest): settle the T3 owner rule and record the crates-io environment by @kixelated in #46
  • feat(t3): exclude T3 from unowned sessions without a T3 client label by @kixelated in #47
  • chore(quest): plan T3 idle-compaction evidence by @kixelated in #48
  • feat(t3): price idle modes by changed cost and name the missing evidence by @kixelated in #50
  • perf!: restore the large-history runtime, cut peak memory and remove --json-full by @kixelated in #49
  • chore(quest): plan T3 removed owners, inherited window and idle guards by @kixelated in #51
  • chore(quest): plan the scan memory cheap wins and per-session fold by @kixelated in #52
  • chore(quest): plan pricing unset Claude windows from documented defaults by @kixelated in #53
  • feat(t3): resolve the inherited Claude window when launch arguments cannot change it by @kixelated in #58
  • feat(compaction): price unset Claude windows from tuned defaults by @kixelated in #59
  • chore(quest): plan per-subagent-type model and cost lines by @kixelated in #55
  • chore(quest): plan auto memory's cost and green/yellow/red verdicts by @kixelated in #60
  • chore(quest): plan a priced-session JSON export for quest usage by @kixelated in #54
  • chore(quest): plan directory windows and the explicit window cap by @kixelated in #61
  • fix(t3)!: bill sessions of removed T3 instances outside current checks by @kixelated in #57
  • fix(compaction)!: tune each recorded directory's unset window under its own sizing by @kixelated in #67
  • chore(quest): record T3 event pruning and wait on the idle-guard scope by @kixelated in #68
  • chore(quest): plan resolving Claude sessions by T3 turn items by @kixelated in #71
  • perf!: cut the scan's redundant memory by @kixelated in #56
  • feat(report)!: color verdicts green, yellow and red and show feature totals by @kixelated in #64
  • fix(compaction): cap explicit windows at the model's context window by @kixelated in #66
  • docs: restate the premise, cut the README, and plan doctor, installers and claim tiers by @kixelated in #62
  • chore(quest): plan gateway windows and landing through the merge queue by @kixelated in #72
  • docs: trim agent instructions and move quest to 93c9c2ef by @kixelated in #70
  • feat: export each session's priced usage with token-max sessions by @kixelated in #69
  • Report each subagent type's model and cost by @kixelated in #63
  • chore(quest): plan sharing subagent attribution with session export by @kixelated in #73
  • feat: report auto memory's total cost and advise against it by @kixelated in #65
  • perf!: analyze each SSH destination on its own host by @kixelated in #78
  • feat: resolve Claude sessions without a T3 thread by turn items by @kixelated in #77
  • Share Codex subagent attribution between the report and session export by @kixelated in #76
  • Price unset gateway windows at the declared context size by @kixelated in #75
  • feat!: remove the --json report by @kixelated in #74
  • ci: add the Pullfrog workflow for PR reviews by @kixelated in #79
  • feat!: sum costs exactly in any order by @kixelated in #80
  • feat: section the report by kind of claim by @kixelated in #82
  • fix: isolate test homes from host config and checkout path length by @kixelated in #81
  • chore: move the repository to the kixel-ai organization by @kixelated in #83
  • ci: run Check on merge queue entries by @kixelated in #88
  • docs: land PRs through the merge queue by @kixelated in #90
  • chore(quest): plan run.ps1 for running doctor once on Windows by @kixelated in #85
  • chore(quest): plan exporting sessions from SSH hosts by @kixelated in #84
  • fix(test): keep test scratch out of the walked checkout by @kixelated in #91
  • feat!: run the report as token-max doctor and offer each change by @kixelated in #86
  • feat: ship prebuilt binaries with cargo-dist by @kixelated in #89
  • feat: serve install and run scripts from max.kixel.ai by @kixelated in #93
  • fix: read Claude usage zeroed in a compacted subagent's carried copy by @kixelated in #95
  • feat!: export sessions from SSH hosts with their repositories by @kixelated in #92
  • perf!: fold the scan per session by @kixelated in #87
  • chore: bump quest to 33f4629 by @moq-bot[bot] in #97
  • chore: bump quest to 0b60de0 by @moq-bot[bot] in #100
  • chore: bump quest to v0.11.0 by @moq-bot[bot] in #101
  • chore: bump quest to v0.11.1 by @moq-bot[bot] in #102
  • chore: add a project icon for T3 Code by @kixelated in #98
  • fix: pair equal compaction parts whose cache charge is unknown on one side by @kixelated in #99
  • chore(quest): plan subscription framing, an effort check, and Fable and subagent guidance by @kixelated in #96
  • chore: bump quest to v0.11.2 by @moq-bot[bot] in #103
  • chore: bump quest to v0.11.3 by @moq-bot[bot] in #108
  • chore: bump quest to v0.12.0 by @moq-bot[bot] in #109
  • feat: cite Anthropic's subagent model guidance beside each subagent line by @kixelated in #106
  • feat: advise subscription users, measured at API prices by @kixelated in #107
  • feat: check each profile's main-session default effort by @kixelated in #104
  • feat(t3): rebuild idle guards from T3's retained event log by @kixelated in #94
  • feat: recommend a main cache TTL while it is unset by @kixelated in #110
  • feat: show Fable's weekly use beside the plan cap by @kixelated in #105
  • chore: bump quest to v0.12.1 by @moq-bot[bot] in #111
  • chore: bump quest to v0.12.2 by @moq-bot[bot] in #113
  • docs(quest): plan doctor's report fixes before publishing 0.0.1 by @kixelated in #112
  • Prove nonuse through unparseable lines by @kixelated in #116
  • fix(doctor): give each setting one row per settings file by @kixelated in #114
  • feat!: cut doctor's default report to one ranked line per setting by @kixelated in #115
  • fix(doctor): prove a shared row's nonuse from any source's whole window by @kixelated in #117
  • chore: bump quest to v0.12.3 by @moq-bot[bot] in #119
  • chore: bump quest to v0.12.4 by @moq-bot[bot] in #120
  • chore: bump quest to v0.13.0 by @moq-bot[bot] in #121
  • chore: bump quest to v0.14.0 by @moq-bot[bot] in #122
  • chore: bump quest to v0.15.0 by @moq-bot[bot] in #123
  • feat(doctor)!: walk the changes biggest saving first, then confirm one list by @kixelated in #118
  • docs(quest): plan the launch: logo, site, README, prettier CLI, descriptions, savings vs defaults by @kixelated in #124
  • docs(quest): color and pad the walk by @kixelated in #125
  • chore: bump quest to v0.15.1 by @moq-bot[bot] in #126
  • docs(quest): group the walk's small changes and offer to remove unused MCP servers by @kixelated in #127
  • docs(logo): add the piggy bank logo, favicon and light/dark lockups by @kixelated in #128
  • feat(doctor): group the walk's small changes by kind and reason by @kixelated in #130
  • feat(doctor): say what each setting and feature does by @kixelated in #131
  • docs(quest): plan doctor --defaults, the report a default user would see by @kixelated in #133
  • fix(doctor): stop offering a deny rule for an unpaired listing delta by @kixelated in #134
  • feat(doctor): show what the settings save against the defaults by @kixelated in #129
  • feat(doctor): offer to remove unused and turned-off MCP servers by @kixelated in #132
  • docs(quest): plan backing up each settings file before doctor writes it by @kixelated in #135
  • feat(doctor): make the walk and report pretty by @kixelated in #136
  • Back up each settings file before doctor writes it by @kixelated in #139
  • feat(doctor): add --defaults, the report as a default user would see it by @kixelated in #137
  • docs(readme): lead with the logo, a recording and the maintainer's results by @kixelated in #138
  • docs(quest): plan extrapolating over unpriced requests and asking a profile's billing by @kixelated in #141
  • docs(quest): call the extrapolation's margin an estimate under its assumption by @kixelated in #142
  • feat(site): a short landing page and README top led by the cache TTL findings by @kixelated in #140
  • chore: bump quest to v0.15.2 by @moq-bot[bot] in #145
  • chore: bump quest to v0.16.0 by @moq-bot[bot] in #146
  • chore: bump quest to v0.16.1 by @moq-bot[bot] in #147
  • docs(quest): finish the site quest now that max.kixel.ai is deployed by @kixelated in #149
  • chore: bump quest to v0.17.2 by @moq-bot[bot] in #148
  • docs(quest): plan auditing several SSH hosts as one by @kixelated in #151
  • docs(quest): plan the 4-host 30-day results snapshot by @kixelated in #152
  • chore: bump quest to v0.20.0 by @moq-bot[bot] in #153
  • feat(doctor): extrapolate over unpriced requests with a ~ and a 95% margin by @kixelated in #143
  • chore: bump quest to v0.22.0 by @moq-bot[bot] in #154
  • feat(doctor): ask whether a Claude profile is on a plan or API billing by @kixelated in #144
  • feat(doctor): extrapolate feature and subagent costs with ~ and a 95% margin by @kixelated in #156
  • Add one total across SSH hosts by @kixelated in #157
  • docs(quest): plan reading shared settings for controls T3 leaves alone by @kixelated in #160
  • feat(doctor): extrapolate the idle compaction replays over unpriced sessions by @kixelated in #159
  • feat(ssh): put this build on each SSH host before running it by @kixelated in #158
  • docs: retake the results snapshot across all four hosts over 30 days by @kixelated in #155
  • chore(quest): deploy the retaken results to max.kixel.ai by @kixelated in #164
  • docs(quest): plan resolving CLAUDE_AUTOCOMPACT_PCT_OVERRIDE by @kixelated in #161
  • chore: bump quest to v0.22.1 by @moq-bot[bot] in #166
  • Apply the walk's changes on SSH hosts by @kixelated in #163
  • feat(doctor): read shared settings for controls T3 leaves alone by @kixelated in #162
  • chore: bump quest to v0.22.2 by @moq-bot[bot] in #168
  • docs: retake the results snapshot with the shared-settings fix by @kixelated in #167
  • feat(doctor): resolve CLAUDE_AUTOCOMPACT_PCT_OVERRIDE in the compaction window check by @kixelated in #165
  • Prepare the token-max 0.0.1 release by @kixelated in #169

New Contributors

  • @kixelated made their first contribution in #1
  • @moq-bot[bot] made their first contribution in #97

Full Changelog: https://github.com/kixel-ai/token-max/commits/v0.0.1