Skip to main content

Changelog

What changed in each release of the DoiT Cloud Intelligence CLI. To get the latest version, run dci update, or see install and update in the CLI guide.

v2.7.5 — September 4, 2026​

New​

  • dci export-datahub-dataset-records exports the records of a DataHub dataset, and the CLI prints that export as the file it is — CSV, or newline-delimited JSON with --format jsonl — instead of an escaped string. See export-datahub-dataset-records.
  • --all exports a whole dataset in one command — dci export-datahub-dataset-records <dataset> --all --output-file records.csv, no dates needed. The API returns one page per request, accepts at most 366 days per request, and requires both time bounds; --all finds the range the dataset actually covers, follows the pages, walks the successive time windows, and merges everything into one export with a single header row (pages can carry different label and metric columns, so the columns are unioned rather than concatenated). Without --all, a partial export now says so on stderr and prints the --page-token to continue from, and a window wider than 366 days is rejected before the request with a pointer to --all.
  • --output-file (-O) writes any command's output to a file instead of the terminal. Pass a path, or a directory (--output-file .) to save under the filename the API suggests.
  • --for-reimport rewrites an exported CSV into the format dci ingest-datahub-events-csv accepts — dropping the export-only batch, source, export_time and updated_by columns and renaming event_id to id — so records can be copied from one dataset to another. See ingest-datahub-events-csv.

Improved​

  • dci list-datahub-datasets shows a display name column when at least one dataset has one, so the name shown in the console and in report results — set with update-datahub-dataset — is visible in the CLI too. Datasets displayed by their name keep the previous three columns.
  • --page-token and --max-results now fail with a clear message on commands that return their whole response in one request, instead of accepting the value and ignoring it. --all still works everywhere and simply notes when there was nothing more to fetch.
  • Leaving the dataset name off a DataHub command now opens the same filter-as-you-type picker the other commands have, instead of a usage error: dci export-datahub-dataset-records, get-datahub-dataset, delete-datahub-dataset, and update-datahub-dataset (there, write the fields and let the picker supply the name: dci update-datahub-dataset description: "New description"). Dataset names also complete on Tab and resolve from a partial name now — previously none of this worked for datasets, because they have no separate ID.
  • The picker and Tab completion reach commands that act on a sub-resource, so dci get-report-config, dci export-cloudflow-flow and their siblings take a name where they previously took only an ID. Commands whose identifier is a number, such as dci get-ticket, no longer offer names they cannot accept.

v2.7.4 — September 4, 2026​

New​

  • dci ingest-datahub-events-csv now uploads a CSV file (or a GZ/ZIP archive of one) to a DataHub dataset from the command line: dci ingest-datahub-events-csv provider: <dataset>, file: @events.csv. The @ attaches the file by name; a file over the API's 30 MB limit, a missing file, or a forgotten @ is reported with a fix before anything is sent. Previously the command failed with an internal marshaling error. See ingest-datahub-events-csv.

v2.7.3 — September 4, 2026​

New​

  • dci <command> --help now shows curated, tested usage examples and a description of each positional argument instead of an example synthesized from the API schema. The first commands covered are the anomaly commands — patch-anomaly, get-anomaly, list-anomalies, and dci anomalies-recent — and the local commands (dci status, dci query, dci open, dci update, dci skill, and others). Every example is checked against the live API before a release ships, so a command line shown in help is one the API accepts. The remaining commands keep their previous help until their examples land.
  • dci commands --json carries the same examples, page notes, and related commands for each command, so agents and scripts can read them without parsing help text.

v2.7.2 — September 4, 2026​

New​

  • DoiT employees no longer need an Anthropic API key for the AI session: once signed in with dci login, dci ai runs on DoiT-provided access — short-lived tokens fetched automatically with your existing DoiT sign-in, refreshed behind the scenes, never written to disk. An explicit key (the ANTHROPIC_API_KEY environment variable, or a key saved with /key set) still takes precedence, and DCI_AI_PROVIDED=off opts out.
  • dci budgets-at-risk and dci anomalies-recent answer two common questions directly — which budgets are currently at risk, and what anomalies appeared recently (bounded with --window) — without hand-building the filter and sort on dci list-budgets or dci list-anomalies each time.
  • --chart=treemap draws each group's share of a report's total as proportional rectangles in the report's colors, with exact shares in the legend. It needs no time axis, so it also charts single-period reports the other chart modes cannot.
  • The AI session's input now shows a command's remaining arguments as faint ghost text — path parameters first, then required body fields — consumed as you type real values. Tab accepts what the ghost offers next: an empty pickable slot opens the name picker (the ghost says so: "enter to pick from a list"), a value slot shows a hint from the command's documented example, and the next body field inserts its name: prefix for you.
  • Array body fields guide and guard: the ghost shows the literal syntax to type (tags: [a, b]), Tab carries the opening bracket, and the line is checked before submission — a plain value where the command expects a list is rejected with the corrected line to copy ("did you mean: tags: [prod, billing]").
  • The bundled agent skill now teaches CloudFlow flow authoring: agents can build or refine flows from a natural-language request with dci build-cloud-flow and dci refine-cloud-flow, or author a flow bundle by hand through the export → edit → dry-run → import loop, with guardrails that keep generated flows honest (clone real flows instead of guessing node parameters, validate with a dry run, land as drafts). Run dci skill update to refresh installed copies.
  • The skill also teaches verifying a flow's behavior, not just its structure: deep-validate a draft with dci test-run-cloudflow-flow --dry-run, run it once as a test (excluded from run history and statistics, approvals never bypassed), and read what every node consumed and produced with dci get-cloudflow-flow-run — so an agent can fix a flow from evidence instead of asking you to check the Console.

Improved​

  • dci build-cloud-flow and dci refine-cloud-flow now answer with a readable result — the created flow's ID, the conversation ID for follow-up refinements, the builder's reply, and the build steps that ran — instead of the raw event stream. Both commands also accept the family-consistent spellings dci build-cloudflow and dci refine-cloudflow.
  • dci list-cloudflows shows each flow's id first — the value dci export-cloudflow-flow and dci refine-cloud-flow need.
  • Table output for humans now lands the rows you came for nearest the prompt: lists sort newest-last, insights rank top-savings-last, and report groups put the largest spenders just above the TOTAL row. Configurable with --output-order, the DCI_OUTPUT_ORDER environment variable, or dci config; JSON/YAML/CSV, agent mode, and piped output keep the classic ordering.
  • When a command's output is taller than the terminal, a dim hint next to the prompt tells you there's more ("↑ N more lines above — scroll up").
  • --help shows more of what the API documents: command categories carry a one-line description, and flags with a documented example show it.
  • dci beta run-report resolves report names like the rest of the CLI: a typed name resolves to its ID, multi-word names work unquoted, and zero arguments open the interactive report picker — in the shell and in the AI session alike.
  • When a command run from the AI session fails on wrong arguments, the result now ends with the command's one-line usage, so you can fix the call without leaving the session.

Fixed​

  • dci get-cloudflow-flow-run prints a failed run in full — status, per-node results, and the failure message — instead of collapsing it into an error envelope that hid exactly the detail needed to diagnose the run.
  • In the AI session: Enter with the completion popup open submits the fully typed command instead of a similarly named suggestion, and completion continues past a group name (/beta run completes to beta run-report). A logged-out session now points at /login instead of telling you to leave the session and run dci login.
  • The sign-in success page's click-to-run suggestion no longer fights the terminal for keystrokes, and Esc cancels pickers and confirmation prompts in the shell, matching the AI session.
  • dci get-report --chart renders the chart next to the prompt instead of above a long table, where it scrolled out of view.

v2.7.1 — August 26, 2026​

New​

  • /key in the AI session manages your Anthropic API key: on its own it shows which key is in use (masked) and where it comes from — including when the ANTHROPIC_API_KEY environment variable overrides a saved key — /key set replaces it through the guided entry, and /key clear removes it. No more hand-editing the settings file.
  • Your first-ever session opens with the full /help text right after the banner, so you don't have to discover the slash-command grammar by accident. Later sessions keep the compact banner.

Improved​

  • dci beta commands complete in the session's popup and run like every other / command, and a bare /beta lists the early-access commands the way you'd type them in the session.
  • /exit now appears in the command popup and in /help alongside /quit (it has always worked as an alias).
  • When the session has no API commands to offer, the banner says so ("API commands appear after /login") instead of just showing a small command count.

Fixed​

  • The interactive session shows the full command set again. In v2.7.0 every session opened with only the built-in commands — no API commands, whatever your sign-in state — and restarting or signing in again couldn't bring them back.
  • /login works inside the session: it hands the browser sign-in your real terminal, then picks up your commands and identity live once you're back. Previously it failed with "cannot open a browser to log in", which left no way back in after a /logout. Signing in or out now also refreshes the banner's role and tenant immediately, and signing in as a different account drops the previous account's customer context and cached names.
  • Changing or clearing the API key while an answer is streaming no longer risks a status line stuck on a spinner.

v2.7.0 — August 26, 2026​

New​

  • Running dci with no arguments at a terminal now opens the interactive AI session — the same one dci ai opens: ask about your cloud costs in plain English, or run any command with a / prefix. In pipes, scripts, and CI, bare dci prints the help screen exactly as before, so nothing changes for automation. Prefer the help screen at your terminal? Run /default help once inside the session; dci --help always prints help.
  • dci beta opens the early-access command surface, starting with asynchronous report execution: dci beta run-report, run-report-config, get-report-operation, get-report-results, and cancel-report-operation. Beta commands use your existing sign-in, customer context, and output formats; idempotency keys are generated for you unless you pass your own. dci commands --beta lists the surface, and accounts not yet enrolled in the early-access program get a clear hint instead of a bare error.
  • Two new chart styles: --chart sparkline draws period totals as a one-line sparkline for the quickest look at the shape, and --chart heatmap draws one row per group and one cell per period, colored by each value's share of the maximum — spot the hot service and the hot month in one glance.
  • When a long AI answer finishes while you're in another window, the session now sends a desktop notification in addition to the terminal bell (in terminals that support notifications and focus reporting), so you can safely switch away during an investigation.

Improved​

  • Budget utilization bars are color-graded: green through amber (past 70%) into red (past 90%), so dci list-budgets shows at-risk budgets at a glance.
  • The AI session got a visual polish: the banner draws the DoiT mark in high resolution (and as a full logo on Kitty, Ghostty, and WezTerm), identity lines are labeled by role and tenant, the spinner is the DoiT mark with a status shimmer, and long waits rotate through the CLI's waiting quips and shift color as they age.
  • Ctrl+L in the session now also jumps back to the latest content along with repainting the frame — one press returns a known-good live view.
  • Mistyping a command at a terminal now also suggests the plain-English alternative (dci ai "…") alongside the usual did-you-mean suggestion.
  • CI systems that allocate a terminal are now detected via the standard CI environment variable and get the script behavior — a pipeline can never hang on an interactive session.

Fixed​

  • The AI session's command popup scrolls: with more matches than fit (say /list), pressing ↓ now moves through all of them — a counter on the highlighted row shows where you are in the list. Previously matches past the sixth row were unreachable.

v2.6.2 — August 25, 2026​

Fixed​

  • dci ai failed to start in v2.6.1 with a flag error ("unable to redefine 'q' shorthand"). The one-shot quiet mode is spelled --quiet — the -q shorthand belongs to another flag and is not used.

v2.6.1 — August 25, 2026​

New​

  • The AI session rings the terminal bell when an answer that ran commands finishes, so you can switch away during a long investigation and still catch the result. /bell turns it off and remembers your choice.
  • One-shot mode can control how much of the investigation you see: dci ai --quiet "question" prints just the answer, and --verbose keeps the full narration even when output is piped to a file.

Improved​

  • In the AI session's command popup, Enter now selects the highlighted command, matching other CLIs (Tab still works too).
  • AI answers label their numbers: cost tables state the currency and usage tables name the metric, with one consistent scale per column. Query results always carry a currency on cost columns — US dollars when the query doesn't choose one.
  • The /mouse choice is remembered across sessions.
  • Starting dci ai without an API key offers the guided key setup right away — type a / command to skip past it, or press Esc to dismiss it.

Fixed​

  • AI session rendering on wide terminals: answers and tables use the full terminal width, and Ctrl+L reliably repaints a frame disturbed by terminal scrolling.
  • dci export-cloudflow-flow writes the complete flow bundle as JSON, so the exported file can actually be imported; import results report validation errors instead of hiding them.

v2.6.0 — August 24, 2026​

New​

  • dci ai opens an interactive AI session in your terminal: ask about your cloud costs in plain English and the AI runs dci commands for you and explains the results — or run any command yourself with a / prefix, with completion and history. One-shot mode answers a single question and exits: dci ai "top 3 cost anomalies this month". AI features use your own Anthropic API key (the session walks you through saving one securely); your questions and the command results the AI reads are sent to Anthropic under your key, and conversations are never stored anywhere but your terminal.
  • Inside the session: /customer shows or switches the customer context (switches the AI makes itself apply to the session only — your saved context is never changed), /model picks the AI model, /export saves the transcript to a file, and Esc cancels a running command or answer. Define your own shortcuts as saved commands.
  • Tune AI answer speed versus depth with the DCI_AI_EFFORT environment variable or the effort setting (low, medium, high; default restores uncapped reasoning). The default is medium, measured to keep answer quality while responding noticeably faster.
  • --rollup <columns> totals report and query rows in the CLI: rows group by the listed result columns, metric columns are summed, and per-period rows collapse into one total per group — no more spreadsheet step for "total per service over the quarter".
  • --search <substring> finds items in any list command by keyword, case-insensitively, across every text field and every page — e.g. dci list-dimensions --search genai.
  • DCI_AI_STATS=1 prints one telemetry line per AI answer (rounds, tool calls, tokens, timing) for scripting and benchmarking.

Improved​

  • The AI session answers multi-step questions faster: independent commands it batches now run concurrently, and long results are presented as a top-10 table plus an aggregated total — ask for the full breakdown when you want every row.

Fixed​

  • Unauthenticated headless runs (CI, scripts) of --help on API commands fail fast with a structured error instead of hanging on a browser login that can never complete.
  • A mistyped flag value no longer surfaces as a destructive-command confirmation prompt.

v2.5.2 — August 21, 2026​

New​

  • dci update updates the CLI in place on every platform, aware of how you installed it: Homebrew, Scoop, and WinGet installs run the package manager's own upgrade after a confirmation; Linux .deb/.rpm installs download, checksum-verify, and install the new package; standalone binaries replace themselves atomically. dci upgrade is now an alias. (Keep it up to date)
  • --chart renders stacked column charts per group by default (largest segments on top), --chart=line keeps the line of period totals, and charts follow your report's color theme. (Table output options)
  • dci login opens DoiT Cloud Intelligence-branded browser pages, and the success page's suggested command is clickable — click it to copy and run your first command. (Authentication)

Fixed​

  • Pivot tables hide rows whose values are all zero.
  • A rejected token (401) now names the API base that rejected it, so pointing at the wrong environment is obvious.
  • DCI_API_BASE_URL applies only to the invocation it is set on — it is never silently persisted into your saved configuration. (Environment variables)
  • The CloudFlow import command correctly requires its idempotency key up front instead of failing on the server.

v2.5.1 — August 20, 2026​

New​

  • Interactive prompts in human mode: a fuzzy picker when a command needs a resource you didn't name, a selection menu when a name matches more than one resource, and a default-Cancel confirmation before destructive commands. Pipes, scripts, and agents never see a prompt. (Resource names and pickers)
  • An interactive table viewer and query builder: explore results full-screen, sort and filter columns, and compose report queries step by step.
  • A spinner shows while an API request is in flight, and a compact one-line hint lists columns hidden from the current table.

Improved​

  • The destructive-command confirmation shows the target's name, owner, and description — you see exactly what you're about to delete. (Destructive commands)
  • dci login got visual polish, budgets render utilization bars, and the update notice is styled instead of plain text.

Fixed​

  • Report tables always pivot for humans — the previous 14-period cutoff that switched wide results to flat rows is gone (--flat still opts out). (Report results)

v2.5.0 — August 19, 2026​

New​

  • --all fetches and merges every page of a paged list response, so you get the complete list without chasing page tokens.
  • --drop-unlabeled-rows excludes report rows whose group labels are all null — only rows with no label at all are dropped, never partially labeled ones.

Improved​

  • dci customer-context set accepts the customer's URL display name as it appears in DoiT Console URLs, alongside domains and customer IDs. (Customer context)
  • Flat report rows omit datetime dimension columns that duplicate the timestamp column.
  • Paged lists tell you when a page token was dropped, and --max-results values beyond the server's cap fail up front with the real limit.

v2.4.0 — August 18, 2026​

New​

  • Event timestamps in table output display in your local timezone; report period columns stay UTC because they label billing buckets, not moments. (Timestamps and timezones)

Fixed​

  • When a resource name can't be resolved, the command falls back to your argument as typed instead of failing the lookup.
  • Table rows come from the command's primary collection, not from sideloaded arrays that happened to be in the response.

v2.3.1 — August 18, 2026​

Fixed​

  • list-tickets renders its curated view for both response shapes the API returns. (list-tickets)

v2.3.0 — August 18, 2026​

New​

  • Curated default table views for eighteen more list commands (reports, budgets, alerts, and others): the most useful columns lead, and ids resolve to names — a report's owner, a role's name — instead of raw identifiers. Machine formats and explicit -C/--fields selections keep the raw fields. (Output formats)

Fixed​

  • Terminal hyperlinks in cells no longer shear table column alignment.

v2.2.0 — August 18, 2026​

Improved​

  • list-insights opens with a curated, savings-ranked view: title-led columns, potential daily savings in USD, easy-win markers, and links to each insight's underlying report. (Insights on the cheat sheet)

v2.1.1 — August 17, 2026​

Fixed​

  • Completing a partial first word offers the matching root commands.
  • A Tab press never triggers a network call or an OAuth login — completion on a cold cache stays local. (Shell completion)

v2.1.0 — August 17, 2026​

New​

  • Use resource names instead of ids: commands resolve names like dci get-report "Monthly costs", and shell completion offers your resource names — dci get-report Mon<TAB> — served from a local cache that refreshes in the background. (Shell completion)
  • Shell completion scripts ship inside every install package.

Improved​

  • The embedded agent guidance covers name resolution, so AI agents address resources by name too.

Fixed​

  • Multi-word resource names work unquoted in dci open, in name resolution, and in completion — and shells that strip quotes still complete quoted names correctly.
  • Sourcing the zsh completion script registers it without extra setup.

v2.0.0 — August 16, 2026​

The CLI's human experience overhaul — see CLI v2.0: before and after for the full gallery of what changed.

New​

  • Report results render for humans: pivoted tables with heatmap shading and trend columns, matching how the console presents a report; machine formats get the same results as structured rows.
  • dci docs prints every documentation entry point, for people and for AI agents. (Documentation for agents)
  • Built-in FinOps workflows and command discovery guidance for AI agents driving the CLI. (Agent mode)

Improved​

  • Requests are validated before they are sent: unknown body fields, malformed path parameters, and invalid non-interactive invocations fail fast with hints instead of opaque server errors.
  • Errors classify by the API's real HTTP status and map to distinct exit codes, and authentication failures come with tailored hints. (Exit codes)

Fixed​

  • The CLI repairs an unusable saved API configuration, recovers from an invalid saved API base, and honors the API base you configure.
  • Customer context resolves from the canonical customer id, and console customers resolve even before any report data exists.
  • Generated pivot totals rows are labeled as such, and successful empty API responses are accepted instead of reported as errors.

For releases before v2.0.0, see the GitHub releases page.

Last updated on