# Retrieval Recovery Regression

> Retrieval Recovery Regression is a paid API for AI agents from www.mahastrategies.com, paid per call via x402, $0.074001/call, status unknown (last checked 2026-10-16).

Compares complete, partial, and missed evidence recovery across a matched, hash-frozen cohort of up to 64 question-paper pairs, reporting paired transitions and a caller-declared complete-recovery drop threshold.

## Facts

- Endpoint: POST https://www.mahastrategies.com/api/v1/context/retrieval/recovery-regression?utm_source=zero.xyz
- Price: $0.074001/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-16
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/retrieval-recovery-regression-4e5b53f3
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_fwBJuVSqBEbz1S7AtZGkv

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability retrieval-recovery-regression-4e5b53f3 -d '<json body>'
```

Example prompt: Run a retrieval recovery regression on this 20-question cohort using my baseline and candidate retrieval objects, pinned to corpus hash 'sha256-abc123...', with public data, and flag if complete-recovery drops more than 150 basis points.

## When to prefer this

Choose this endpoint when you need to compare evidence-ID recovery (complete, partial, missed) between two retrieval configurations on a reproducible, hash-frozen cohort of up to 64 question-paper pairs, and want a caller-declared threshold test for complete-recovery regression. It is not designed for answer quality evaluation or causal inference — use it strictly for descriptive overlap analysis of retrieval outputs.

## Known failure modes

- Mismatched array lengths between questions, paperIds, and evidence (validation error)
- corpusHash mismatch causing cohort freeze failure
- maxCompleteDropBps out of range 0–10000
- questions or paperIds arrays exceeding 64-item limit
- Invalid dataClass value (must be 'public' or 'synthetic')
- Missing required baseline or candidate objects
- Payment failure via x402 protocol

## How this service works

Compare complete, partial, and missed evidence recovery on one matched, hash-frozen cohort of up to 64 questions and papers. Report paired transitions and a caller-declared complete-recovery drop threshold. Descriptive evidence-ID overlap, not answer quality or causal inference. New modular adapter: retained result recovery for 24 hours, minimal receipt metadata for 90 days. Existing live route unchanged.

## Output

Returns paired transition counts between complete, partial, and missed evidence recovery states, descriptive evidence-ID overlap statistics, and a flag indicating whether the caller-declared complete-recovery drop threshold (in basis points) was exceeded. Results are retained for 24 hours; minimal receipt metadata is kept for 90 days.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "baseline": {
   "type": "object"
  },
  "evidence": {
   "type": "array",
   "maxItems": 64,
   "minItems": 1
  },
  "paperIds": {
   "type": "array",
   "maxItems": 64,
   "minItems": 1
  },
  "candidate": {
   "type": "object"
  },
  "dataClass": {
   "enum": [
    "public",
    "synthetic"
   ],
   "type": "string"
  },
  "questions": {
   "type": "array",
   "maxItems": 64,
   "minItems": 1
  },
  "corpusHash": {
   "type": "string",
   "maxLength": 71
  },
  "maxCompleteDropBps": {
   "type": "integer",
   "maximum": 10000,
   "minimum": 0
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/retrieval-recovery-regression-4e5b53f3/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from www.mahastrategies.com](https://www.zero.xyz/host/www.mahastrategies.com/llms.txt)