Add cost estimationAdd dollar cost estimation across 4 models

[aether-png]

What does this change?
Adds dollar cost estimation across 4 models (Sonnet, Haiku, Gemini Flash, GPT-4o)
to the preflight output. One new line shows estimated cost before you run:

Dollar cost: ~$0.0330 (sonnet) | ~$0.0088 (haiku) | ~$0.0007 (gemini) | ~$0.0225 (gpt-4o)

Assumes output = 2x input tokens. Adds scripts/lib/cost.js + tests.
Also fixes estimate.test.js to use spawnSync over execSync
(resolves Windows EPERM issue in sandboxed environments).

Does npm test pass?

• Yes (confirmed locally, spawnSync fix resolves Windows sandbox issue)

Does this violate any design principles?

• No — zero new dependencies, output is proportional, heuristics disclosed

[ZonatedCord]

Thanks for the PR — the direction is right, but there's a critical bug
before this can merge.

Bug #1 (blocking): estimate.js:141 uses taskResult.tokens instead of
totalTokens. This means the cost shown reflects only the task description —
not the system prompt, CLAUDE.md, open files, or conversation turns. That's
exactly what Mimir is designed to capture. The estimate would be drastically
understated and misleading.

Fix: replace estimateAllModelCosts(taskResult.tokens) →
estimateAllModelCosts(totalTokens).

Also blocking: --output json should include costInfo — otherwise JSON
consumers get incomplete data.

Minor (can follow in separate PR): hardcoded prices, output token assumption
(2x input overstimates read-heavy tasks), GPT-4o/Gemini inclusion in a
Claude-specific tool is a design question worth discussing.

Worth merging now, separate PR: the spawnSync fix — clean improvement, no
issues. Fix the two blocking issues and this is good to go.

[aether-png]

Fixed both blocking issues:

• estimateAllModelCosts now uses totalTokens (not taskResult.tokens)
• costInfo included in --output json payload

ready for re-review.

[ZonatedCord]

Hey @aether-png — thanks for the PR and for pushing the fix so fast. The core of this feature is solid. Here's my full review.

---

✅ What works

• cost.js — clean module, zero dependencies, tests pass. Good separation of concerns.
• execSync → spawnSync in tests: real improvement, more robust.
• costInfo in the JSON output: correct and useful.

---

🚨 Blockers (must fix before merge)

1. require('./lib/presets') — file missing (estimate.js:13)
scripts/lib/presets.js is not included in the PR. Every call to estimate.js crashes immediately on import.

2. estimatePresetCosts not exported (estimate.js:12, cost.js:62)
cost.js imports estimatePresetCosts from estimate.js but it's not in the exports. Guaranteed crash when using --preset.

3. DEBTOKEN PREFLIGHT instead of MIMIR PREFLIGHT (estimate.js:211)
Your fork's name leaked into the main codebase. This breaks 3 tests in estimate.test.js (lines 20, 41, 66 — all assert.match(..., /MIMIR PREFLIGHT/)). Simple find/replace fix.

---

🐛 Logic bug

4. Hardcoded values in --preset branch (estimate.js:114–118)
$0.49, $0.13, $0.005, $0.485 are hardcoded. The presetCosts variable is calculated at line 107 but then never used. If prices update, this branch silently stays wrong.

---

🧹 Design / cleanup

5. --preset feature is half-baked and wasn't in the PR description
It depends on presets.js (missing), estimatePresetCosts (missing), and has hardcoded values. Incomplete feature that shouldn't land in this PR. I'd suggest one of two paths:

• Cut it entirely from this PR and open a separate one once the dependencies are ready.
• Complete it with presets.js included and hardcoded values removed.

6. Self-referential README credits (README.md:118)

"This cost estimate support is a fork of github.com/ZonatedCord/Mimir"

We are the ZonatedCord/Mimir repo. This line doesn't make sense here. Remove it (or I can attribute the feature to you in the changelog/readme in a way that actually makes sense).

---

The spawnSync fix + cost.js module are mergeable as-is once the blockers above are resolved. Happy to review a v3 quickly.

[aether-png]

addressed all review feedback:

• reverted banner to MIMIR PREFLIGHT
• removed all --preset code (presets.js, estimatePresetCosts, TASK_PRESETS)
• cost calculation now uses totalTokens correctly
• removed Credits subsection from README
• costInfo remains in --output json
• npm test passes

ready for v3 review.

[ZonatedCord]

Hey — almost there. One thing: looks like some "Debtoken" references from your fork leaked into the diff. Totally understandable, classic fork situation.

Things to fix before merge:

README.md

• Title: # Debtoken → # Mimir
• All install paths: ~/.claude/debtoken → ~/.claude/mimir (same for Gemini/Codex sections)
• Example output: ⚡ DEBTOKEN PREFLIGHT → ⚡ MIMIR PREFLIGHT

tests/estimate.test.js

• assert.match(outHigh, /DEBTOKEN SPLIT/ → /MIMIR SPLIT/

.gitignore

• AGENTS.md and .agents/ look like personal config from your setup — remove them from this PR

Overall though — solid work. The in-process test rewrite alone was beyond what I asked for. Really appreciate the effort you're putting into this.

[aether-png]

I've removed the remaining "Debtoken" references as requested:

• README.md: Reverted the title to # Mimir, changed all install/verify paths to mimir instead of debtoken (across Claude Code, Gemini, and Codex sections), and fixed the example output preflight header.
• tests/estimate.test.js: Updated the assertion from /DEBTOKEN SPLIT/ back to /MIMIR SPLIT/.
• .gitignore: Removed the AGENTS.md and .agents/ lines.
  Ready for another look