yevice

1. Scope

yevice is a Rust CLI (yevice) that generates a cost model from IaC templates and evaluates monthly cost estimates from that model. Subcommands: generate, eval, compare, sensitivity, validate, simulate, optimize, update-pricing, diagram. Supported IaC formats: CloudFormation (YAML/JSON), Terraform (HCL), Cloudflare Wrangler (wrangler.toml). Repository: https://github.com/rabe1028/yevice

---

2. Check installation

which yevice

If not found, install from a yevice checkout:

./scripts/install.sh                 # default ~/.cargo/bin/yevice
./scripts/install.sh --highs         # also enable HiGHS MILP backend

Or clone and install:

git clone https://github.com/rabe1028/yevice.git
cd yevice && ./scripts/install.sh

Or directly via cargo:

cargo install --git https://github.com/rabe1028/yevice.git yevice-cli

---

3. Typical workflows

A. Estimate cost from IaC (baseline flow)

yevice generate --template path/to/template.yaml --name app --output ./out
yevice eval ./out/app.cost_model.json --params ./out/app.usage.yaml
yevice eval ./out/app.cost_model.json --params ./out/app.usage.yaml \
  --display-currency JPY --exchange-rate USD:JPY=155

• --input-format accepts: auto (default), cfn, tf, wrangler
• --strict treats parse failures as fatal (recommended in CI)

Run A.5 before eval — usage.yaml is an empty template after generate.

---

A.5 Interactive parameter completion (Claude-driven)

After generate, three files are produced:

• ./out/app.cost_model.json — expression AST and resource graph
• ./out/app.schema.json — JSON Schema describing each parameter
• ./out/app.usage.yaml — parameter file (empty template, must be filled)

Always inspect <name>.schema.json first. It defines for each parameter: name, type (number / integer / string), unit (e.g. requests/month, GB, ratio), description, default (if any), and enum (if any).

Claude follows this procedure to fill usage.yaml interactively:

1. Read <name>.schema.json and enumerate all properties.
2. For each property:
   • If default is present, show it as the candidate and confirm with the user.
   • If no default, ask the user for a value (AskUserQuestion), presenting the unit and description together.
   • For wide-range numeric parameters (e.g. requests, storage_gb), preview 3–4 log-scale examples: 1e6, 1e7, 1e8, 1e9.
   • If enum is present, list all choices.
3. Write the collected values into usage.yaml (Edit or Write tool).
4. Run yevice eval --params ./out/app.usage.yaml.

Filled values persist in the file, so subsequent eval runs are non-interactive.

Important: When usage.yaml is empty or contains placeholder values (e.g. null, 0, "REPLACE_ME"), do NOT silently fall back to defaults — ask the user. Cost estimates with wrong assumptions are worse than no estimate.

Unit conventions (schema is authoritative; these are common defaults):

Unit	Meaning
requests	per-month request count
storage_gb	GB-month
egress_gb	GB out per month
duration_ms	per-invocation duration
concurrency, replicas	integer count
cache_hit_ratio	0.0–1.0 fraction

---

B. Sensitivity analysis

Run A.5 first to ensure usage.yaml has values before invoking sensitivity.

yevice sensitivity ./out/app.cost_model.json \
  --params ./out/app.usage.yaml \
  --var requests --min 1e6 --max 1e9 --steps 10

---

C. Cost optimization (MILP)

Enumeration solver (no extra dependency):

yevice optimize ./out/app.cost_model.json \
  --params ./out/app.usage.yaml \
  --budget 500

HiGHS solver (requires --highs install):

yevice optimize ./out/app.cost_model.json \
  --params ./out/app.usage.yaml \
  --budget 500 --solver highs

---

D. Compare scenarios

yevice compare \
  --baseline ./out/v1.cost_model.json --baseline-params ./out/v1.usage.yaml \
  --candidate ./out/v2.cost_model.json --candidate-params ./out/v2.usage.yaml

---

E. Refresh pricing catalog

yevice update-pricing --region ap-northeast-1

---

4. File formats

Kind	Description
Input	CFN YAML/JSON, TF HCL, wrangler.toml
<name>.cost_model.json	Expression AST + resource graph
<name>.schema.json	JSON Schema for parameter file. Read this before filling usage.yaml
<name>.usage.yaml	Parameter file (must be filled before eval / sensitivity)
pricing-data/<service>.json	AWS Price List catalog

---

5. Currency and FX

The default catalog currency is USD. To display in another currency:

--display-currency JPY --exchange-rate USD:JPY=155.0

When resources span multiple pricing currencies, the total is labeled mixed. Set --exchange-rate to resolve to a single currency.

---

6. Parse policy

• Lenient (default): best-effort parse; unrecognized resources are skipped with a diagnostic.
• --strict: any parse failure is fatal; recommended for CI.

Diagnostics have the shape:

IacParseDiagnostic { code, severity, source, location, message }

---

7. Picking the right subcommand

Question	Subcommand
PR cost delta	compare
What if requests 10x?	sensitivity
Cheapest SKU mix	optimize
Capacity / quota check	validate
Time-varying peak load	simulate --profile
Just need a number	generate then eval (run A.5 first)

---

8. Troubleshooting

• Linker error at build — install a C linker (gcc / clang).
• HiGHS not available — reinstall with ./scripts/install.sh --highs.
• AWS price API rate limit — retry after a moment; use cached pricing-data/ if available.
• TF parse failures — check the supported Terraform provider/service list in the repo README.
• Mixed-currency totals — set --exchange-rate to normalise to one currency.
• usage.yaml empty / eval result is $0 — run interactive completion (A.5) to fill in parameter values.

---

9. Project home

For installation, design rationale, and release notes, see https://github.com/rabe1028/yevice.