PRODUCTION USE CASE
A legal operations team compares a master services agreement with a security addendum and pricing schedule. The parser preserves page numbers, headings, tables, and footnotes. The model receives selected sections and produces a clause-by-clause deviation report.
Production recipe
| Production choice | Recommendation | Why |
|---|---|---|
| API | POST /v1/chat/completions | OpenAI-compatible server-side request |
| Primary model | gemini-2.5-pro | Gemini 2.5 Pro is the first route for large mixed document sets and cross-document comparison. |
| Fallback model | gpt-5.6-terra | Use Terra for text-only sets already reduced to the relevant sections. Escalate only the unresolved high-risk clauses to Sol or counsel. |
| Escalation model | gpt-5.6-sol | Use only when the primary route fails the defined quality or complexity boundary |
| Output contract | Validated JSON | A clause-level deviation table with document and page citations for every row. |
A legal operations team compares a master services agreement with a security addendum and pricing schedule. The parser preserves page numbers, headings, tables, and footnotes. The model receives selected sections and produces a clause-by-clause deviation report.
Every deviation must cite the exact document and page. OCR confidence is checked before the model call; a bad scan is reprocessed instead of asking the model to guess.
Send the request from your server. Replace the example content and placeholder tool schema with data and tools from your application.
curl https://app.llmfly.ai/v1/chat/completions \
-H "Authorization: Bearer $LLMFLY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-2.5-pro",
"messages": [
{"role": "system", "content": "Compare the supplied contract sections. Return a deviation table with issue, baseline clause, changed clause, risk, document ID, and page. Never infer missing text."},
{"role": "user", "content": "Compare liability, security, renewal, and pricing terms across these documents."}
],
"response_format": {"type": "json_object"}
}'Gemini 2.5 Pro is the first route for large mixed document sets and cross-document comparison.
Use Terra for text-only sets already reduced to the relevant sections. Escalate only the unresolved high-risk clauses to Sol or counsel.
| Metric | Pass condition |
|---|---|
| Evidence accuracy | Every material answer points to the correct page, section, table, or footnote |
| Cross-section recall | Cross-section questions include all necessary supporting passages |
| Table and footnote handling | Values and qualifications in tables and footnotes survive parsing correctly |
| Cost per document | Parsing, retrieval, generation, and verification stay within the per-document budget |
A clause-level deviation table with document and page citations for every row.
Record the model ID, request ID, token usage, retries, validation result, and final disposition for every production run.
Only when the tested route, task, latency, and cost support it; retrieval often still helps.
Preserve page identifiers and require material answers to point to supporting passages.
Evaluate OCR quality separately before judging the language model.
Compare the primary and fallback models with the same requests, tools, and validation rules.