Turbo Mode

Context calls are cached in Redis per scopeId + task pair. Task matching is normalized for small casing, quote, punctuation, and whitespace differences, but different intent still creates a new cache entry. Hits are optimized for low latency (often under 20ms p50). Misses run the full retrieval path (on the order of ~200ms) and populate the cache — so marketing comparisons should pair “fast” with cached /v1/context, not with every retrieval path or with self-hosted vector infrastructure you still have to run.

Modes

| Mode | Behavior | P50 latency | |---|---|---| | fast (default) | Read cache first. Miss → compute + store | <20ms cached / ~200ms miss | | fresh | Always bypass cache. Full retrieval every time. | ~200ms |

How to use

Pass mode as a query parameter or header:

bash
# query parameter
POST /v1/context?mode=fast
POST /v1/context?mode=fresh

# or as a header
X-Memcone-Mode: fast
X-Memcone-Mode: fresh
# query parameter
POST /v1/context?mode=fast
POST /v1/context?mode=fresh

# or as a header
X-Memcone-Mode: fast
X-Memcone-Mode: fresh
typescript
// fast (default)
const res = await fetch('https://api.memcone.com/v1/context', {
  method: 'POST',
  headers: { 'Authorization': `Bearer ${key}`, 'Content-Type': 'application/json' },
  body: JSON.stringify({ scopeId, task }),
})

// fresh — bypasses cache
const res = await fetch('https://api.memcone.com/v1/context?mode=fresh', {
  method: 'POST',
  headers: { 'Authorization': `Bearer ${key}`, 'Content-Type': 'application/json' },
  body: JSON.stringify({ scopeId, task }),
})
// fast (default)
const res = await fetch('https://api.memcone.com/v1/context', {
  method: 'POST',
  headers: { 'Authorization': `Bearer ${key}`, 'Content-Type': 'application/json' },
  body: JSON.stringify({ scopeId, task }),
})

// fresh — bypasses cache
const res = await fetch('https://api.memcone.com/v1/context?mode=fresh', {
  method: 'POST',
  headers: { 'Authorization': `Bearer ${key}`, 'Content-Type': 'application/json' },
  body: JSON.stringify({ scopeId, task }),
})

Cache invalidation

The cache is invalidated automatically when remember changes memory for the same scope. You never need to manually clear it.

text
POST /v1/remember { scopeId: "user_123", ... }
→ if new memory was stored or an active belief changed, Memcone bumps the version counter for "user_123"
→ all prior cache entries for "user_123" are immediately stale
→ next /v1/context call recomputes and caches fresh result
POST /v1/remember { scopeId: "user_123", ... }
→ if new memory was stored or an active belief changed, Memcone bumps the version counter for "user_123"
→ all prior cache entries for "user_123" are immediately stale
→ next /v1/context call recomputes and caches fresh result

Reading cache status

Every /v1/context response includes:

json
{
  "result": "...",
  "cache_hit": true,
  "tokens_saved": 84
}
{
  "result": "...",
  "cache_hit": true,
  "tokens_saved": 84
}

Response headers also expose it:

text
X-Memcone-Cache: HIT
X-Memcone-Latency-Ms: 14
X-Memcone-Tokens-Saved: 84
X-Memcone-Cache: HIT
X-Memcone-Latency-Ms: 14
X-Memcone-Tokens-Saved: 84

When to use fresh

  • After a series of remember calls when you want the most current state immediately
  • In background jobs where latency doesn't matter
  • When debugging — fresh always shows the live state of memory
PreviousGET /v1/usage
Next Debug mode