burnwatch 0.5.1 → 0.7.0

package/CHANGELOG.md

@@ -5,11 +5,60 @@ All notable changes to burnwatch will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.7.0] - 2026-03-25
+
+### Added
+
+- **Service probing system**: New extensible `probes.ts` module that auto-detects plan tiers, usage, and billing data from service APIs. Adding a new service probe is a single function — the system supports N services. Probes for 9 of 14 services:
+  - **Scrapfly**: Plan name + credit usage (high confidence — skips plan selection)
+  - **Vercel**: Plan tier from team/user API (high confidence)
+  - **Supabase**: Plan tier from Management API (high confidence, requires PAT)
+  - **Anthropic**: Current month USD spend from Admin API cost report (medium — shows spend, still asks plan)
+  - **OpenAI**: Token usage from Admin API (medium)
+  - **Stripe**: Balance and processing volume (medium)
+  - **Browserbase**: Session count and browser hours (medium)
+  - **Upstash**: Database discovery (low — validates key)
+  - **PostHog**: Organization discovery (low — validates key)
+- **Tiered discovery in interview**: Init now follows API → key → ask → hedge:
+  1. Find API key (env vars, global config)
+  2. If key found + probe exists → hit the API, show what we found
+  3. High confidence → "Detected: Pro ($100/mo, 1M credits). Correct? [Y/n]"
+  4. Medium confidence → show usage data, then ask plan
+  5. No key → show plan list, ask for key after
+  6. Budget always set
+- **Auto-configure probing**: Non-TTY mode (Claude Code) also probes APIs when keys are available, auto-matching detected plans instead of always using defaults.
+
+### Changed
+
+- **Interview no longer requires `autoDetectPlan` in registry**: Any service with a probe in `PROBES` map is automatically probed when a key is available. Adding a service-specific probe is the only requirement.
+- **Key detection runs before plan selection for all services**: Previously only LIVE-tier services checked for keys. Now any service with a probe gets key detection first.
+
+## [0.6.0] - 2026-03-24
+
+### Added
+
+- **Allowance tracking for credit-pool services**: Services like Scrapfly that sell a fixed credit pool (e.g., Pro $100/mo = 1M credits) now track unit consumption against the plan allowance, not just dollar spend. The brief shows `↳ 850K/1M credits (85%) ⚠️` alongside the dollar line. This is the distinction between "budget" (what you pay) and "spend metric" (what you consume).
+- **Allowance data in spend snapshots**: `SpendSnapshot.allowance` provides `used`, `included`, `unitName`, and `percent` for credit-pool services with LIVE connectors.
+- **PlanTier `includedUnits` and `unitName`**: Registry plans can now declare how many units are included (e.g., `"includedUnits": 1000000, "unitName": "credits"`). Init automatically sets `TrackedService.allowance` from plan selection.
+
+### Changed
+
+- **Scrapfly registry updated to real pricing**: Discovery $30/200K, Pro $100/1M (default), Startup $250/2.5M, Enterprise $500/5.5M. Previously had incorrect plan names and prices.
+- **Scrapfly connector returns unit data**: `BillingResult` now includes `unitsUsed`, `unitsTotal`, and `unitName` so the brief can show credit consumption alongside dollar spend.
+- **Allowance-aware status labels**: Credit-pool services show "750K credits left" or "⚠️ 125% of 1M credits used" instead of generic budget percentages.
+
+## [0.5.2] - 2026-03-24
+
+### Changed
+
+- **Zero manual commands after init**: Every detected service is now fully configured during init - plan, tier, and budget. No wall of `burnwatch add` commands. Registry now includes `suggestedBudget` for all usage-based plans (Anthropic $100, OpenAI $100, Stripe $50, Google Gemini $50, Voyage AI $20, AWS $50). Flat-rate services get budget = plan cost. Free tiers get $0.
+- **Interactive interview uses suggested budgets as defaults**: Usage plans show `Monthly budget [$100]: $` instead of a blank prompt. Press Enter to accept.
+
 ## [0.5.1] - 2026-03-24
 
 ### Fixed
 
-- **Init actually works now**: `process.stdin.isTTY` was `undefined` in Claude Code and many terminal environments, so the interactive interview never ran. Init now has three modes: (1) TTY detected = full interactive interview, (2) no TTY = smart auto-configure with defaults, (3) `--non-interactive` = minimal CI mode. The auto-configure path picks each service's default plan, detects API keys from env vars and global config, sets budget = plan cost for flat plans and $0 for usage-based, and only shows `burnwatch add` commands for the handful of usage-based services that genuinely need a human-set budget.
+- **Init actually works now**: `process.stdin.isTTY` was `undefined` in Claude Code and many terminal environments, so the interactive interview never ran. Init now has three modes: (1) TTY detected = full interactive interview, (2) no TTY = smart auto-configure with defaults, (3) `--non-interactive` = minimal CI mode.
 
 ## [0.5.0] - 2026-03-24
 
@@ -72,6 +121,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Snapshot system for delta computation across sessions
 - Claude Code skills: `/spend` (on-demand brief), `/setup-burnwatch` (guided onboarding)
 
+[0.7.0]: https://github.com/RaleighSF/burnwatch/compare/v0.6.0...v0.7.0
+[0.6.0]: https://github.com/RaleighSF/burnwatch/compare/v0.5.2...v0.6.0
+[0.5.2]: https://github.com/RaleighSF/burnwatch/compare/v0.5.1...v0.5.2
 [0.5.1]: https://github.com/RaleighSF/burnwatch/compare/v0.5.0...v0.5.1
 [0.5.0]: https://github.com/RaleighSF/burnwatch/compare/v0.4.3...v0.5.0
 [0.4.3]: https://github.com/RaleighSF/burnwatch/compare/v0.4.2...v0.4.3