repospend 0.0.1 → 0.0.3

package/CHANGELOG.md

@@ -0,0 +1,24 @@
+# Changelog
+
+All notable changes to RepoSpend will be documented in this file.
+
+## 0.0.3
+
+- Add initial Claude Code support for local JSONL session transcripts, including projects, timestamps, models, token usage when available, and source status.
+- Add source filtering across dashboard data and CLI summaries with `--source all|codex|claude`.
+- Add bundled Claude pricing coverage, cache write input pricing, and Claude family pricing fallback for dated model IDs.
+- Refresh the dashboard experience with source-aware views, Settings improvements, and fictional demo screenshots.
+- Document supported Codex and Claude Code data sources, pricing assumptions, privacy boundaries, security reporting, and release notes.
+- Remove the roadmap document from `docs/`.
+
+## 0.0.2
+
+- Add GitHub CI and npm publish workflow support.
+- Improve npm package metadata for discovery and release publishing.
+- Keep RepoSpend focused on local-first AI coding usage visibility.
+
+## 0.0.1
+
+- Initial public release.
+- Add Codex-first local usage dashboard.
+- Add CLI entry point for running RepoSpend locally.

package/README.md

@@ -1,37 +1,21 @@
 # RepoSpend
 
-RepoSpend is a local dashboard for seeing where your AI coding usage is going.
+[![npm version](https://img.shields.io/npm/v/repospend)](https://www.npmjs.com/package/repospend)
+[![npm downloads](https://img.shields.io/npm/dm/repospend)](https://www.npmjs.com/package/repospend)
+[![license](https://img.shields.io/npm/l/repospend)](./LICENSE)
+[![node](https://img.shields.io/node/v/repospend)](https://www.npmjs.com/package/repospend)
+[![CI](https://github.com/mehmetdemircs/RepoSpend/actions/workflows/ci.yml/badge.svg)](https://github.com/mehmetdemircs/RepoSpend/actions/workflows/ci.yml)
 
-It reads your local Codex data, groups work by Git repository, and shows the repos,
-sessions, models, commands, and days that are driving token usage and
-API-equivalent cost.
+RepoSpend is a local-first dashboard for seeing which repositories are using the
+most AI coding tokens.
 
-`0.0.1` is an early Codex-first release.
+It reads supported local Codex and Claude Code files in read-only mode, groups
+sessions by Git repo, and shows token usage, cost estimates, models, sessions,
+and agent friction.
 
-## Why RepoSpend?
+No login. No telemetry. No prompt uploads.
 
-AI coding usage usually gets shown by date, model, or raw session. RepoSpend starts
-with the thing most developers actually care about: the repo.
-
-If Codex records work from both of these paths:
-
-```text
-/Users/elrond/dev/RivendellRecords
-/Users/elrond/dev/RivendellRecords/apps/web
-```
-
-RepoSpend walks up to the Git root and shows them together as one
-`RivendellRecords` project.
-
-That makes it easier to answer questions like:
-
-- Which repo is using the most tokens?
-- Which sessions were unusually expensive?
-- How much came from cached input, output, or reasoning tokens?
-- Are command failures or repeated tool issues adding friction?
-- Which models and days are driving the total?
-
-## Run It
+## Quick Start
 
 You can run RepoSpend without installing it globally:
 
@@ -39,6 +23,8 @@ You can run RepoSpend without installing it globally:
 npx repospend
 ```
 
+Then open the local dashboard URL printed in your terminal.
+
 Or install it once:
 
 ```bash
@@ -53,42 +39,152 @@ prints the dashboard URL. By default it runs at
 RepoSpend requires Node.js `20` or newer. It uses `better-sqlite3`, so npm may
 install a native SQLite package for your platform.
 
-## Commands
+## Good For
 
-Most of the time you only need:
+- finding which repo is burning the most AI coding tokens
+- reviewing expensive or unusual coding-agent sessions
+- comparing Codex and Claude Code usage locally
+- spotting repeated command failures and agent friction
+- exporting usage data for your own analysis
 
-```bash
-repospend
-```
+## Screenshots
 
-There are also a few terminal-friendly commands:
+Screenshots use fictional Middle-earth demo data. The Lord of the Rings themed
+repo names, sessions, prompts, token counts, and costs are intentional; no private
+repository data is shown.
 
-```bash
-repospend scan
-repospend by-repo
-repospend by-day
-repospend by-hour
-repospend by-model
-repospend by-app
-repospend export --format json
-repospend export --format csv
+### Overview
+
+![RepoSpend overview dashboard with fictional Middle-earth usage data](docs/screenshots/dashboard-overview.png)
+
+The main dashboard summarizes AI coding tokens, API-equivalent cost, top repos,
+cache reuse, file edits, and command issue rate.
+
+### Repositories
+
+![RepoSpend repositories table with fictional repo usage](docs/screenshots/repos-view.png)
+
+The repos view compares spend, tokens, sessions, cache hit rate, file edits, and
+token intensity across projects.
+
+### Repository Detail
+
+![RepoSpend repository detail for one-ring-infra](docs/screenshots/repo-detail.png)
+
+Repo detail explains why a project stands out, including cost concentration,
+warnings, token shape, sessions, and command signals.
+
+### Sessions
+
+![RepoSpend sessions table with fictional session titles](docs/screenshots/sessions-view.png)
+
+The sessions view makes individual AI coding runs searchable and sortable by
+repo, tool, model, outcome, cost, tokens, and activity.
+
+### Session Detail
+
+![RepoSpend session detail for a fictional palantir event stream fix](docs/screenshots/session-detail.png)
+
+Session detail shows the shape of one run: cost, tokens, model, file edits,
+commands, highlights, and issues to inspect.
+
+### Agent Friction
+
+![RepoSpend agent friction screen with fictional command issue signals](docs/screenshots/agent-friction.png)
+
+Agent Friction separates blocking command failures from harmless shell exits so
+high-token troubleshooting is easier to review.
+
+## What It Shows
+
+RepoSpend helps you break down local AI coding usage by:
+
+- repo
+- session
+- day and hour
+- model
+- source/tool and app/surface, where detectable
+- token type
+- estimated API-equivalent cost
+
+If Codex records work from both of these paths:
+
+```text
+/Users/elrond/dev/RivendellRecords
+/Users/elrond/dev/RivendellRecords/apps/web
 ```
 
-Use `REPOSPEND_NO_OPEN=1 repospend` if you want the URL printed without opening a
-browser.
+RepoSpend walks up to the Git root and shows them together as one
+`RivendellRecords` project.
+
+## Supported Tools
+
+| Tool | Status | Notes |
+|---|---|---|
+| Codex | Most complete support | Tokens, models, sessions, repo grouping, command friction |
+| Claude Code | Initial support | Sessions, projects, models, timestamps, tokens when available |
+| RTK | Optional/local | Shown only when local RTK data exists |
+
+RepoSpend started as a Codex-first release. Claude Code support is newer and
+depends on what your local Claude Code files include.
+
+## Privacy
+
+RepoSpend is local-first:
+
+- no login
+- no telemetry
+- no prompt uploads
+- no changes to Codex or Claude Code files
+
+Everything the dashboard shows comes from files already on your machine.
+
+## Cost Estimates, Not Invoices
+
+RepoSpend shows **API-equivalent cost**.
+
+That means it estimates cost from local token counts and the pricing assumptions
+stored in RepoSpend. It is useful for comparing repos and sessions, but it is not
+an invoice.
+
+Your actual cost may be different because of subscriptions, credits, included
+usage, account-level terms, provider changes, or other billing details. If you use
+Codex or Claude Code through a subscription, read the number as "what this token
+usage would roughly cost at API-style rates."
+
+Claude Code cost may show as unknown when local files do not include token counts
+or a model name. RepoSpend shows those sessions with unknown tokens/cost rather
+than guessing.
 
 ## What It Reads
 
-RepoSpend reads Codex data from:
+RepoSpend only reads local files. It does not edit Codex, Claude Code, or RTK
+data.
+
+Codex data:
 
 ```text
 ~/.codex/state_5.sqlite
 ~/.codex/sessions
 ```
 
-Those files are opened read-only. RepoSpend does not modify Codex data.
+Claude Code data:
 
-RepoSpend stores its own settings here:
+```text
+~/.claude/projects
+~/.config/claude/projects
+~/Library/Application Support/Claude/local-agent-mode-sessions
+~/.config/Claude/local-agent-mode-sessions
+```
+
+RepoSpend also checks `~/.claude/history.jsonl` for source status, but does not
+import history-only entries into usage analytics because they do not contain
+reliable token/model data.
+
+Claude Code transcript files can contain prompt text, tool output, and file
+contents. RepoSpend keeps all scanning local.
+
+RepoSpend-owned settings:
 
 ```text
 ~/.repospend/pricing.json
@@ -96,31 +192,54 @@ RepoSpend stores its own settings here:
 ```
 
 The Settings page includes a reset action for RepoSpend-owned files under
-`~/.repospend/`. It does not delete or edit anything under `~/.codex`.
+`~/.repospend/`. It does not delete or edit anything under `~/.codex` or
+`~/.claude`.
 
-## About Cost
+## Commands
 
-RepoSpend shows **API-equivalent cost**.
+Most users only need:
 
-That means it estimates cost from local token counts and the pricing assumptions
-stored in RepoSpend. It is useful for comparing repos and sessions, but it is not
-an invoice.
+```bash
+repospend
+```
 
-Your actual cost may be different because of subscriptions, credits, included
-usage, account-level terms, provider changes, or other billing details. If you use
-Codex through a subscription, read the number as "what this token usage would
-roughly cost at API-style rates."
+There are also a few terminal-friendly commands:
 
-## Privacy
+```bash
+repospend scan
+repospend by-repo
+repospend by-day
+repospend by-hour
+repospend by-model
+repospend by-app
+repospend export --format json
+repospend export --format csv
+```
 
-RepoSpend is local-first:
+Most commands also accept a simple source filter:
 
-- no login
-- no telemetry
-- no prompt uploads
-- no changes to Codex files
+```bash
+repospend by-repo --source codex
+repospend by-repo --source claude
+repospend by-repo --source all
+```
 
-Everything the dashboard shows comes from files already on your machine.
+Use `REPOSPEND_NO_OPEN=1 repospend` if you want the URL printed without opening a
+browser.
+
+## Current Limits
+
+- Codex support is still the most complete path for token and command-friction
+  analysis.
+- Claude Code support is initial: sessions, projects, timestamps, models, and
+  token usage are shown when present in local JSONL files.
+- Claude Code sessions without local token details are shown with unknown
+  tokens/cost.
+- Cost estimates do not represent subscription billing, credits, regional
+  pricing, or account-specific terms.
+- Budget alerts are not active yet.
+- Some older sessions may not include full token, command, or prompt details.
+- RTK analytics appear only when local `rtk` data is available.
 
 ## Develop
 
@@ -144,14 +263,17 @@ pnpm build
 [http://127.0.0.1:4318](http://127.0.0.1:4318) and the dashboard on
 [http://127.0.0.1:2005](http://127.0.0.1:2005), with `/api` proxied locally.
 
-## Current Limits
+## Maintainers
 
-- Codex is the first supported data source.
-- Cost is estimated, not billed.
-- Budget alerts are not active yet.
-- Some older sessions may not include full token, command, or prompt details.
-- Unknown models show token totals with unknown cost instead of guessing.
-- RTK analytics appear only when local `rtk` data is available.
+Publishing notes are in [docs/PUBLISHING.md](docs/PUBLISHING.md).
+
+## Security
+
+Please see [SECURITY.md](SECURITY.md) for reporting security issues.
+
+## Changelog
+
+Please see [CHANGELOG.md](CHANGELOG.md) for release notes.
 
 ## License
 

package/SECURITY.md

@@ -0,0 +1,65 @@
+# Security Policy
+
+RepoSpend is an early-stage, local-first developer tool. It is designed to help you inspect AI coding usage data on your own machine, not to collect or upload that data.
+
+## Supported Versions
+
+While RepoSpend is early-stage, only the latest published npm version is actively supported for security fixes.
+
+Please upgrade to the latest version before reporting a vulnerability when possible.
+
+## Reporting a Vulnerability
+
+Please do not open a public GitHub issue for suspected vulnerabilities.
+
+Please contact the maintainer privately via GitHub. If the repository later lists a dedicated private security reporting method or maintainer contact, use that instead.
+
+When reporting, include:
+
+- the RepoSpend version
+- your operating system and Node.js version
+- a clear description of the issue
+- minimal steps to reproduce it
+- whether the issue involves local files, logs, prompts, or usage data
+
+## Security Expectations
+
+RepoSpend should:
+
+- run locally by default
+- read supported AI coding tool data without modifying it
+- avoid uploading prompts, logs, local files, telemetry, or local usage data by default
+- keep security-sensitive changes small and reviewable
+
+RepoSpend is not guaranteed to be free of vulnerabilities. Please use care when running any local tool that reads files from your machine.
+
+## Local-First Data Handling
+
+RepoSpend inspects local usage data, including Codex-related local files and logs when available. Some source files may contain prompt text, tool output, file paths, repository names, or other sensitive details.
+
+Before sharing logs, screenshots, exports, or reproduction steps, remove private data such as prompts, file contents, access tokens, usernames, repo names, customer names, and local paths you do not want to disclose.
+
+## What Not to Include in Public Issues
+
+Please do not include the following in public GitHub issues:
+
+- prompts or conversation logs
+- access tokens, API keys, or secrets
+- private repository names or file paths
+- customer, employer, or personal data
+- raw local usage files that may contain sensitive content
+- details for an unpatched security issue
+
+Use a private maintainer contact for anything security-sensitive.
+
+## Maintainer Response Expectations
+
+Maintainers will try to acknowledge valid vulnerability reports promptly, but response times may vary because RepoSpend is a small open-source project.
+
+The expected process is:
+
+- review the report privately
+- ask follow-up questions if needed
+- prepare a fix when the issue is reproducible and in scope
+- publish a release or mitigation notes when appropriate
+- credit reporters when requested and safe to do so