fathom-cli 0.1.1 → 0.1.3

package/README.md

@@ -15,8 +15,12 @@ npm install -g fathom-cli
 ## Quick Start
 
 ```bash
-# Estimate a feature in seconds
-fathom estimate "React dashboard with OAuth" --complexity L
+# Quick estimate — "how much will this cost?"
+# AI scores complexity automatically when ANTHROPIC_API_KEY is set
+fathom estimate "React dashboard with OAuth"
+
+# Generate a full spec sheet — "break this down into work items"
+fathom intake "Users say the calendar is slow and Zoom links are missing" --project myapp
 
 # Analyze a full spec with multiple features
 fathom analyze spec.md --project myapp
@@ -38,9 +42,9 @@ fathom velocity --project myapp
 
 | Command | Description |
 |---------|-------------|
-| `fathom estimate <desc>` | Quick single-feature token + cost estimate |
+| `fathom estimate <desc>` | Quick cost estimate — AI scores complexity, returns tokens + cost |
+| `fathom intake <text>` | Generate spec sheet — extracts work items, estimates each, outputs markdown |
 | `fathom analyze <spec>` | Parse spec, estimate all features, generate registry |
-| `fathom intake <text>` | Extract work items from raw feedback via Claude API |
 | `fathom track` | Import Claude Code sessions, auto-tag to features |
 | `fathom reconcile` | Compare estimates vs actuals per feature |
 | `fathom calibrate` | Show drift and suggest profile adjustments |
@@ -53,16 +57,28 @@ fathom velocity --project myapp
 
 ## How It Works
 
-### 1. Estimate (Before Work)
+### 1. Estimate vs Intake (Before Work)
 
-Write a spec with `### Feature:` headings, or just describe a feature:
+**`estimate`** — quick cost number from a description:
 
 ```bash
-fathom estimate "Build a real-time chat system with WebSocket support" --complexity L
-# → 205,865 tokens, $2.51 estimated cost, recommends Sonnet
+fathom estimate "Build a real-time chat system with WebSocket support"
+# AI analyzes → L complexity, 205K tokens, $2.51, recommends Sonnet
+# Use --complexity M to override, --no-ai for heuristic-only scoring
 ```
 
-Estimates use a calibrated model: base tokens per task type × complexity multiplier × overhead (context resend, tool calls, thinking, error recovery, planning, review).
+**`intake`** — full spec sheet from raw feedback:
+
+```bash
+fathom intake "Users need real-time chat and the Zoom integration is broken" --project myapp
+# Extracts discrete work items, classifies each by complexity/priority,
+# estimates tokens + cost per item, attaches velocity benchmarks,
+# outputs structured markdown spec ready for implementation
+```
+
+Use `estimate` when you want a quick number. Use `intake` when you want a plan.
+
+Both use a calibrated model: base tokens per task type × complexity multiplier × overhead (context resend, tool calls, thinking, error recovery, planning, review).
 
 ### 2. Track (During Work)
 
@@ -127,7 +143,7 @@ Requires `ANTHROPIC_API_KEY` environment variable.
 
 | Variable | Required | Purpose |
 |----------|----------|---------|
-| `ANTHROPIC_API_KEY` | For `intake` only | Claude API for work item extraction |
+| `ANTHROPIC_API_KEY` | For `estimate` + `intake` | AI complexity scoring and work item extraction |
 | `CONVEX_URL` | Optional | Enable dashboard sync |
 
 ## Spec Format

package/dist/data/remote-manifest.json

@@ -0,0 +1,78 @@
+{
+  "version": 1,
+  "updated": "2026-03-05",
+  "models": [
+    {
+      "id": "claude-opus-4-6",
+      "provider": "anthropic",
+      "name": "Claude Opus 4.6",
+      "input_per_mtok": 15.0,
+      "output_per_mtok": 75.0,
+      "cache_read_per_mtok": 1.5,
+      "batch_discount": 0.5,
+      "tier": "frontier"
+    },
+    {
+      "id": "claude-sonnet-4-6",
+      "provider": "anthropic",
+      "name": "Claude Sonnet 4.6",
+      "input_per_mtok": 3.0,
+      "output_per_mtok": 15.0,
+      "cache_read_per_mtok": 0.3,
+      "batch_discount": 0.5,
+      "tier": "balanced"
+    },
+    {
+      "id": "claude-haiku-4-5",
+      "provider": "anthropic",
+      "name": "Claude Haiku 4.5",
+      "input_per_mtok": 1.0,
+      "output_per_mtok": 5.0,
+      "cache_read_per_mtok": 0.1,
+      "batch_discount": 0.5,
+      "tier": "speed"
+    }
+  ],
+  "recommendations": {
+    "architecture": {
+      "model": "claude-opus-4-6",
+      "reason": "Deep reasoning, trade-off analysis"
+    },
+    "complex-implementation": {
+      "model": "claude-sonnet-4-6",
+      "reason": "Best cost/capability for multi-file coding"
+    },
+    "standard-crud": {
+      "model": "claude-sonnet-4-6",
+      "reason": "Reliable, fast, cost-effective"
+    },
+    "scaffolding": {
+      "model": "claude-haiku-4-5",
+      "reason": "Speed + cost over reasoning depth"
+    },
+    "documentation": {
+      "model": "claude-haiku-4-5",
+      "reason": "Straightforward generation"
+    },
+    "test-generation": {
+      "model": "claude-sonnet-4-6",
+      "reason": "Needs code logic understanding"
+    },
+    "code-review": {
+      "model": "claude-opus-4-6",
+      "reason": "Benefits from deeper reasoning"
+    },
+    "debugging": {
+      "model": "claude-opus-4-6",
+      "reason": "Benefits from deeper reasoning"
+    },
+    "refactoring-architectural": {
+      "model": "claude-opus-4-6",
+      "reason": "Needs system-wide understanding"
+    },
+    "refactoring-mechanical": {
+      "model": "claude-sonnet-4-6",
+      "reason": "Pattern application"
+    }
+  }
+}