Skip to content
Olympusx AIOlympusx AIOlympusx AI documentation

Limits

This page gathers every numeric limit a person or an administrator meets in Olympusx AI. Each row says whether the value is fixed or can be changed, and by whom, and links to the page that explains it. Limits on API calls are also on API limits.

“Fixed” means the value is part of the release. “Settings” means a platform administrator changes it on Platform › Settings. “Configuration” means it is set in the installation’s configuration.

Gateway and API

Section titled “Gateway and API”
Limit Value Can change Explained in
A model call that sends nothing 300 s (30 to 3,600) Settings API limits
One whole call through the gateway, the Playground, Ask or assistant chat 1,800 s (60 to 14,400) Settings API limits
A request body through the gateway 1 MiB Fixed API limits
A request to an external model 32 MiB Fixed API limits
Public assistant chat, per tenant 60 requests a minute Configuration Agents and assistants over the API
Public assistant chat message 4,000 characters; request 64 KiB Fixed Agents and assistants over the API
End-user reference (user, endUserRef, u) 200 characters, no @ Fixed Chat completions

Keys and budgets

Section titled “Keys and budgets”
Limit Value Can change Explained in
Application key lifetime 1 to 3,650 days, or none; the console proposes 90 Per key Keys and budgets
Developer key lifetime 1 to 365 days, 90 by default Per tenant, on the Code Assistant page Developer keys
Active developer keys per person 1 to 50, 5 by default Per tenant, on the Code Assistant page Developer keys
Agent run key 1 hour Fixed Keys, budgets and metering
Budget periods Daily or monthly Per budget Keys and budgets
Budget evaluation Every few seconds Fixed Keys, budgets and metering
Budget flags in the console 50% and 80% of the limit Fixed Keys and budgets

Agents, tools and skills

Section titled “Agents, tools and skills”
Limit Value Can change Explained in
Tool calls in one run 20 Per agent, or per run over the API Agents, tools and skills
Model calls in one run 25 Per agent, or per run Agents, tools and skills
Tokens in one run 200,000 Per agent, or per run Agents, tools and skills
One run, wall clock 5 minutes Fixed Runs
Collaborators per agent 12 Fixed Collaborators
Hand-offs from the agent a person spoke to 2 Fixed Collaborators
Agent name 1 to 80 characters Fixed Agents
Routing key 280 characters Fixed Agents
Temperature 0 to 2 Per agent Create and edit an agent
Passages from knowledge 6 by default, at most 20 Per agent Create and edit an agent
Agent YAML imported 256 KiB Fixed Agents
One tool call 60 s Configuration Tools
One tool result the model reads 32 KiB; longer is cut Configuration Tools
Listing a tool server’s tools 8 s Fixed Tools
Tool server credential header name 64 characters Fixed Tools
Tools one skill narrows to 64 Fixed Skills

Approvals

Section titled “Approvals”
Limit Value Can change Explained in
A held call waits 24 hours, then expires unrun Fixed Human approval
Arguments shown to an approver The first 16,000 bytes Fixed Approvals
A resumed run after approval 5 minutes Fixed Approvals

Assistants

Section titled “Assistants”
Limit Value Can change Explained in
Assistant name 1 to 80 characters Fixed Assistants
One message 4,000 characters Fixed Chatting with an assistant
Passages from knowledge 6 by default, at most 20 Per assistant Assistants
Share links per assistant 1; a new one revokes the old Fixed Publish and share

Knowledge

Section titled “Knowledge”
Limit Value Can change Explained in
One uploaded or fetched document 16 MiB Fixed Collections
Chunks per document 2,000; a larger document fails Fixed Collections
Chunk size 1,000 characters by default, 100 to 8,000 Per pipeline Pipelines and connectors
Chunk overlap 100 characters by default Per pipeline Pipelines and connectors
Pipeline schedule Hourly, daily, weekly or manual Per pipeline Pipelines and connectors
Passages retrieved per question 6 by default, at most 20 Per agent or assistant Knowledge and grounding
Relevance floor 0.5 (0 to 1; 0 turns it off) Settings, and per embedding model Knowledge and grounding
Test search on a collection 5 by default, at most 50 Per search Collections
Vector database sizes Small 2 CPU, 2 GiB, 10 GiB; medium 4 CPU, 4 GiB, 20 GiB; large 8 CPU, 16 GiB, 64 GiB; custom up to 128 CPU Per database Vector databases
Vector database backup retention 30 days by default, 1 to 3,650 Per database Vector databases
One collection graph build 30 minutes Fixed The collection graph
Graph rebuild After 20 s without changes, at most every 5 minutes Fixed The collection graph

Code Assistant

Section titled “Code Assistant”
Limit Value Can change Explained in
Code search matches 8 by default, 1 to 20 Per search Repository view
Code search query 4,000 characters Fixed Code tools
What uses this: steps 1 to 3, 1 by default Per search Repository view
What uses this: elements in one answer 500 Fixed Code tools
find_usages path and symbol 1,000 and 300 characters Fixed Code tools
One codebase tool call 30 s Fixed Code tools
Ask: turns kept 12; older ones drop Fixed Repository view
Ask: code excerpts behind an answer 8 Fixed Repository view
Ask: one question 16,000 characters Fixed Repository view
Largest file indexed 512 KiB by default, up to 16 MiB Per code source Code sources
Files in one source listing 50,000; a sync past it fails Fixed Code sources
Groups and repositories one source names 100 Fixed Code sources
Include and exclude patterns 2,000 characters each Fixed Code sources
Daily sync 02:00 UTC by default Per code source Code sources
One sync 30 minutes Fixed Code sources
Chunks per file 2,000 Fixed Code sources
Device name on a developer key 64 characters Fixed Developer keys
Merge request diff read 1 MiB of patches; every changed file is still listed Fixed Merge request review
Merge request review size 20 batches of 24,000 characters; a larger merge request is reviewed in part Fixed Merge request review
Comments per batch 15 Fixed Merge request review
Reading and scanning the diff 10 minutes Fixed Merge request review
One review 1 hour Fixed Merge request review
Reviews listed 200 Fixed Merge request review
Webhook delivery 2 MiB Fixed Webhooks

Security

Section titled “Security”
Limit Value Can change Explained in
One model call of an assessment, review or investigation 600 s (60 to 3,600) Settings Settings and limits
Files in one code assessment 3,000; the rest are skipped and reported Fixed Code assessments
Assessment batch 24,000 characters, up to 25 findings each Fixed Code assessments
One assessment 6 hours Fixed Code assessments
Assessments listed 100; findings listed 500 Fixed Findings
Dismissal reason 1,000 characters Fixed Findings
Investigation rounds 4 Fixed Investigations
Queries per round 3 Fixed Investigations
Rows per query 200 Fixed Investigations
Characters read per round 14,000 Fixed Investigations
Findings per investigation 10, each with up to 40 rows of evidence Fixed Investigations
Proposed actions per investigation 3 Fixed Actions and approvals
One investigation 30 minutes Fixed Investigations
Investigation question 2,000 characters Fixed Investigations
Investigation window 7 days by default, at most 90 Per investigation Investigations
Hunts 3 presets; each runs at most once a day On or off per tenant Hunts
Hunt window Since its last run, at most 90 days; first run 24 hours Fixed Hunts
Log source query 200 rows, 30 s Fixed Log sources
Log source name 80 characters Fixed Log sources
Alerts opened per delivery 5 Fixed Alert integration
Window around an alert 2 hours before, 30 minutes after Fixed Alert integration
Alert delivery 1 MiB Fixed Webhooks

Guardrails, review and records

Section titled “Guardrails, review and records”
Limit Value Can change Explained in
Review capture 1 in N requests, N from 0 (off) to 1,000,000 Per tenant Guardrails
A captured answer 32 KiB, then truncated Fixed Review queue
Audit export 200,000 rows; a marker row says when it was cut Fixed Audit
Scores export 200,000 rows Fixed Scores
Searches on runs, traces and conversations 200 characters Fixed Runs
Observability time range 92 days Fixed Observability overview
System health history 30 days Fixed System health
Retention windows Keep forever by default; captured gateway turns 30 days Platform admin, over the management API Data retention
Retention sweep Once a day Fixed Data retention
A Settings change takes effect Within 30 seconds Fixed Settings

People and sign-in

Section titled “People and sign-in”
Limit Value Can change Explained in
Tenant roles 5 Fixed Roles
Invitation link 12 hours In the identity service Users and roles
Temporary password 20 characters Fixed Users and roles
Password policy, built-in identity service 12 characters with upper case, lower case, a digit and a symbol, not the username Configuration Identity and sign-in
Sign-in session, built-in identity service 30 minutes idle, 10 hours at most; access tokens last 15 minutes Configuration Identity and sign-in
Display name 120 characters Fixed Profile and preferences

Models

Section titled “Models”
Limit Value Can change Explained in
One model import 4 hours Fixed Model registry
Replicas offered by the deploy wizard Up to 32, within the tenant’s capacity Per deployment Deploy a model
Model server log shown 200 lines by default, at most 1,000 Per view Model details
Model display name 120 characters Fixed The model catalog
Section titled “Related”