Sema

Appearance

Cost Tracking & Budgets ​

Usage Tracking ​

llm/last-usage ​

Get token usage from the most recent LLM call.

scheme
(llm/last-usage)   ; => {:prompt-tokens 42 :completion-tokens 15 ...}

llm/session-usage ​

Get cumulative usage across all LLM calls in the current session.

scheme
(llm/session-usage)

llm/reset-usage ​

Reset session usage counters.

scheme
(llm/reset-usage)

Pricing Sources ​

Sema tracks LLM costs using pricing data from multiple sources, checked in this order:

  1. Custom pricing — set via (llm/set-pricing "model" input output), always wins
  2. Dynamic pricing — fetched from llm-prices.com during (llm/auto-configure), cached locally at ~/.sema/pricing-cache.json
  3. Built-in estimates — hardcoded fallback table (may be outdated)
  4. Unknown — if no source matches, cost tracking returns nil and budget enforcement is best-effort

Dynamic pricing is fetched with a short timeout (2s) and failures are silently ignored. The language works fully offline — the cache persists between sessions.

llm/pricing-status ​

Check which pricing source is active and when it was last updated.

scheme
(llm/pricing-status)
; => {:source fetched :updated-at "2025-10-10"}
; or {:source hardcoded} if no dynamic pricing is available

Budget Enforcement ​

Note: If pricing is unknown for a model (not in any source), budget enforcement operates in best-effort mode — the call proceeds with a one-time warning. Use (llm/set-pricing) to set pricing for unlisted models.

llm/set-budget ​

Set a spending limit (in dollars) for the session. LLM calls that would exceed the budget will fail.

scheme
(llm/set-budget 1.00)   ; set $1.00 spending limit

llm/budget-remaining ​

Check current budget status.

scheme
(llm/budget-remaining)   ; => {:limit 1.0 :spent 0.05 :remaining 0.95}

llm/clear-budget ​

Remove the spending limit.

scheme
(llm/clear-budget)

llm/set-pricing ​

Set custom pricing for a model (overrides both dynamic and built-in pricing). Costs are per million tokens.

scheme
(llm/set-pricing "my-model" 1.0 3.0)   ; $1.00/M input, $3.00/M output

Batch & Parallel ​

llm/batch ​

Send multiple prompts concurrently and collect all results.

scheme
(llm/batch ["Translate 'hello' to French"
            "Translate 'hello' to Spanish"
            "Translate 'hello' to German"])

llm/pmap ​

Map a function over items, sending all resulting prompts in parallel.

scheme
(llm/pmap
  (fn (word) (format "Define: ~a" word))
  '("serendipity" "ephemeral" "ubiquitous")
  {:max-tokens 50})