bazr-cli

Terminal client for BAZR, the Solana meme aftermarket.

BAZR only looks at tokens that already graduated from a launchpad. A relic score is a weighted summary of survival signals with every input shown; it is not a prediction of price or revival. There is no trending feed, no new-launch feed and no sniping in this tool, by design.

+-[o]-- RELIC TAG ----------------------------------------------------------+
| EXAMPLE                                                                   |
| So11111111111111111111111111111111111111112                               |
+---------------------------------------------------------------------------+
| SCORE       47 / 100  [#####.....]                                        |
| VERDICT     unclear                                                       |
| OBSERVED    3 of 5 axes observed  (70% of the weight)                     |
| GRADUATED   2025-11-02                                                    |
| SCORED      2026-03-19 07:00:00Z  cache hit, age 120s                     |
+---------------------------------------------------------------------------+

AXIS BREAKDOWN
AXIS               SCORE  WEIGHT     SHARE  CONTRIB  METER
-----------------  -----  ------  --------  -------  ------------
Holder dispersion     62     25%       36%     22.1  [######....]
LP residual           30     25%       36%     10.7  [###.......]
Dev wallet state      --     20%  excluded  no data  [no data   ]
Floor shape           55     20%       29%     15.7  [######....]
Social afterglow      --     10%  excluded  no data  [no data   ]

Install

This package is not published to the npm registry yet. npm install -g bazr-cli and npx bazr-cli both fail with E404 today. Build it from source instead:

git clone https://github.com/BazrMarket/bazr-sdk.git
cd bazr-sdk/packages/cli
npm install
npm run build
npm test

node dist/bazr.js --help

npm install here resolves @bazr/sdk from the committed lockfile as a link to the sibling ../sdk-ts, so no registry and no separate linking step are involved. npm test builds both bundles before it runs.

To put bazr on your PATH without a registry:

cd bazr-sdk/packages/cli && npm link
bazr --help

When a release does happen: @bazr/sdk first, then bazr-cli

bazr-cli depends on @bazr/sdk by registry range (^0.1.0), not by path. Publishing the CLI first uploads a package nobody can install, and the error names the SDK rather than the CLI, so it is easy to go looking in the wrong place:

npm error code E404
npm error 404 Not Found - GET https://registry.npmjs.org/@bazr%2fsdk - Not found
npm error 404
npm error 404  '@bazr/sdk@^0.1.0' is not in this registry.

So: publish packages/sdk-ts first, then packages/cli. @bazr/sdk is a scoped package and therefore private by default -- the publishConfig.access of public already in its manifest is what makes step one work at all.

Commands

| Command | What it prints | | --- | --- | | bazr relic <mint> | Relic tag, five-axis breakdown, labels, sources | | bazr tags <mint> | Labels only | | bazr stalls | Stall ranking with wins and losses side by side | | bazr stall <owner> | One stall and everything on its table | | bazr crate list | Crates currently assembled | | bazr crate show <id> | One crate and its components | | bazr haggle | Route simulation over existing liquidity | | bazr stats | Counters the service reports | | bazr health | Whether the service answers |

bazr relic So11111111111111111111111111111111111111112
bazr relic <mint> --refresh
bazr stalls --sort record --limit 20
bazr crate show 3
bazr haggle --in <mint> --out <mint> --amount 1000000 --slippage-bps 100
bazr tags <mint> --json

Options

| Option | Meaning | | --- | --- | | --api <url> | API base URL. Defaults to $BAZR_API, then http://localhost:8030 | | --json, -j | Print the validated API response as JSON | | --timeout <ms> | Per-attempt timeout. Default 90000 for relic and tags, 10000 for everything else | | --retries <n> | Total attempts including the first (default 3) | | --color / --no-color | Force colour on or off. Auto-detected otherwise | | --debug | Print the stack trace when something fails |

Colour is switched off automatically when stdout is not a TTY, when NO_COLOR is set, and when TERM=dumb. Everything the CLI prints reads correctly with colour stripped: there are no emoji and no glyph-only status markers, only words and O / X.

The timeout default is per endpoint because the endpoints are not comparable. /health and /stall answer in well under a second. Scoring a mint the service has not seen walks its holder pages and asks several upstreams first -- measured against production, /relic/{mint} took 28.8s cold and /relic/{mint}/tags 47.7s. A slow request says so on stderr rather than sitting silent, and so does every retry, so a slow run can be told apart from a stuck one. Both notices go to stderr; --json on stdout stays clean.

What the output promises

• An axis that could not be observed prints -- and no data, never 0. It is excluded from the weighting and the remaining weights are re-normalised. The footer states how many axes were observed and how much of the weight that covered.
• The contribution column adds up. The sum is printed next to the score the API reported, so a mismatch is visible rather than hidden.
• Stall records print wins and losses as separate raw counts. There is no win-rate column: a rate alone lets a bad record hide behind its denominator.
• Labels are printed as reported, including low-confidence alerts. Rug and bundle traces are not filtered out.
• haggle is a simulation over existing liquidity. BAZR runs no order book and no AMM of its own; the quote names the router it used. High price impact gets its own warning block, not a footnote.
• The API disclaimer is printed on every command.

Exit codes

| Code | Meaning | | --- | --- | | 0 | Success | | 1 | The request failed (network, HTTP error, contract violation) | | 2 | Usage error (unknown command or option, missing argument) | | 70 | bazr stopped before producing a result -- a bug in bazr itself |

70 exists because the worst failure a CLI can have is one that reports success. If any promise in the command never settles, Node empties its event loop and leaves with no output and a status no script would read as a failure. That shipped once: an awaited backoff timer had been told not to hold the process open, and bazr relic died mid-retry on every cache miss while exiting as though nothing were wrong. The timer is fixed and covered by tests that run this binary as a real process; 70 is the guard that makes any future recurrence loud instead of silent.

Failures print one sentence plus hints, not a stack trace. Add --debug when you want the stack.

X  Cannot reach http://localhost:8030/relic/<mint>: fetch failed (ECONNREFUSED)
   Is the BAZR service running and reachable?
   Point the client somewhere else with --api <url> or the BAZR_API env var.

Development

The CLI depends on @bazr/sdk from the same repository. The full loop from a clean checkout, all of which passes:

cd packages/sdk-ts && npm install && npm run typecheck && npm run build && npm test
cd ../cli        && npm install && npm run typecheck && npm run build && npm test

The committed package-lock.json records @bazr/sdk as a link to ../sdk-ts, so a plain npm install sets it up. If that link ever needs re-establishing by hand, npm run dev:link installs ../sdk-ts into node_modules without touching package.json, so the manifest keeps the registry range ^0.1.0 rather than a local file: path.

npm test builds both packages before it runs. test/spawned.test.ts executes dist/bazr.js as a real child process, because that is the only place some failures are visible: called in-process, the CLI borrows the test runner's event loop, and a command that would strand a bare node dist/bazr.js returns normally. Every other suite here passed while bazr relic printed nothing at all against the live service. Testing the shipped binary against a stale @bazr/sdk/dist would reintroduce the same blind spot, which is why the build covers both.

Gates

A glyph scan guards what this CLI is allowed to render. It ships with the repository and covers both packages, so anyone can run it at any time.

# from packages/cli -- glyph scan over sdk-ts and cli
npm run gate:emoji:selftest   # control group. Run this first.
npm run gate:emoji

Attach the real output, not a tick:

selftest ok=8 fail=0 verdict=PASS

scanned=50 excluded=1 unreadable=0
  EXCLUDED (declared) .../packages/cli/scripts/gate-emoji.mjs
hits=0
verdict=PASS

scanned= is counted from the same file list the scan reads, so it cannot drift from what was actually inspected. scanned=0 is SELF-FAIL, not PASS: not looking and finding nothing produce identical output otherwise.

Excluded files are named on every run rather than dropped quietly. They are the detectors themselves and the test fixtures, which necessarily contain the words and glyphs being searched for. gate:emoji:selftest is the control group: it checks that the detector fires on seeded violations and stays quiet on clean source, because a check that always fails passes an audit just as easily as one that always succeeds.

License

MIT