Files
Julius Brussee 82864c8f4c feat: ship Caveman 2.0.0 — engine, proxy, SDKs, unified CLI
Single-commit port of the Caveman 2 tree onto the original repo history
(candidate caveman-2@88878a4 applied over main@3098342). No unrelated-history
merge; original history untouched.

Preserved v1 surfaces:
- skills/caveman/SKILL.md byte-for-byte, plus plugin mirror and caveman.skill zip
- legacy CI (Node 18/20/22 installer + Python 3.11 suites) alongside engine-ci
- cli/install.js compatibility shim forwarding to bin/install.js
- .codex/config.toml ships both hooks and codex_hooks for older Codex (#617)
- README Trendshift + Skills.sh badges and local star-history asset
- v1.10.0-pinned curl installs; benchmarks/tmp-starcharts ignore rules

Fixed for public same-repo release:
- SECURITY.md vulnerability reporting points at this repository
- release-binaries.yml publishes to this repo via job-scoped GITHUB_TOKEN
  (cross-repo PAT removed); binary-release environment still gates signing
- packages/cli/PUBLISHING.md rewritten for packages/ layout and same-repo flow
- profile-contract and provider-catalog lanes skip lane-scope gates on
  initial import so the launch PR can gate on their real validation steps
- ANNOUNCEMENT.md network claim corrected: compression is local; requests
  still go to the provider you configured

Caveman 2 stays opt-in: bin-v1.0.0 binaries and @caveman-ai/cli publish
separately after this lands and soaks.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BLvb5QniwYjmcawyssQNw9
2026-08-11 16:46:10 +02:00

56 lines
2.0 KiB
Go

// Package sampling holds the one deterministic sampler every service shares.
//
// The construction is tenant-first, length-prefixed SHA-256 mapped onto the top
// 53 bits — the same mapping experiments.AssignBucket and the gateway's
// count-token baseline already used. Length prefixing matters: without it the
// key lists ("ab","c") and ("a","bc") would hash identically, and one tenant
// could steer another tenant's selection by choosing a request id. Nothing in
// the request body is hashed, so a caller can never influence its own selection.
//
// The point of sharing it is that the gateway and the worker must agree: the
// same (org, project, monitor, request) is sampled the same way in both, on
// every run, forever. Changing this hash silently reshuffles every population.
package sampling
import (
"crypto/sha256"
"encoding/binary"
"hash"
)
// Fraction maps the ordered key list onto a stable value in [0, 1). The same
// keys always yield the same fraction; keys are length-prefixed so no two
// distinct key lists can collide by concatenation.
func Fraction(keys ...string) float64 {
h := sha256.New()
for _, key := range keys {
writePart(h, key)
}
sum := h.Sum(nil)
u := binary.BigEndian.Uint64(sum[:8])
return float64(u>>11) / float64(1<<53)
}
// Sampled reports whether the key list falls inside the given rate. rate <= 0
// selects nothing and rate >= 1 selects everything; a NaN rate fails closed
// (every comparison against NaN is false, so nothing is selected). Callers are
// responsible for rejecting an incomplete key scope before calling — an empty
// key still hashes, so a missing tenant id would otherwise sample as if it were
// a real, shared scope.
func Sampled(rate float64, keys ...string) bool {
if rate <= 0 {
return false
}
if rate >= 1 {
return true
}
return Fraction(keys...) < rate
}
func writePart(h hash.Hash, value string) {
var n [8]byte
binary.BigEndian.PutUint64(n[:], uint64(len(value)))
_, _ = h.Write(n[:])
_, _ = h.Write([]byte(value))
}