Recent index activity
Service availability
Availability is based on real endpoint responses, not simulated metrics.
Generate once. Reuse with proof.
Connect a video backend, import an existing clip or generate a new one. Make local edits without paying for another model run, then verify every version.
Sign in with a managed account to use Video Studio.
Video projects
Your video backend
Choose a configured backend or register a connector. Customer API keys go to the server and are not saved in the browser.
Connect an API provider
Loading the operator-approved connection policy…
Have a video already?
Skip generation entirely. Register an existing MP4 as a verifiable source asset (up to 25 MiB in the dashboard).
Version history
Revise without regenerating
Change metadata and reuse the exact MP4. Captions or trims invoke local FFmpeg, not an AI model.
Verify & prepare an NFT
Select a version to inspect its provenance.
Knowledge search
Search live builds, published disk segments and your accepted funded shards. Retrieve retained sources independently of previews.
Search indexesAll indexes
Choose which indexes the backend searches. Unchecked indexes are not queried. Your funded indexes are visible only to your account.
Expand to load the indexes visible to this session.
Start typing to search. Guest access is available.
Speak to the knowledge layer
Signed-in conversations are durable, searchable through the private ECAI term index and can be grouped into projects. Guest chat remains an isolated session. Replies depend on the configured backend and are not cryptographic proofs.
Index jobs
Enqueue curated sources, inspect persisted jobs and follow execution events.
Try the complete indexing flow
Use the bundled JSONL file to test inspection, queueing, live progress, search and source retrieval. No external source download is needed. Quick Index still requires the node's IPFS service and index worker.
- Load & inspectReview the detected JSONL structure.
- IndexConfirm using Add to index queue.
- Follow the jobOpen Execution history for live events.
- Search & retrieveFind orbital orchard and open its source.
Loading only selects and inspects the sample. Queueing is a separate explicit action.
What is in the sample?
Orbital orchard — irrigation check; Kinetic lighthouse — battery check; Cedar observatory — sensor check. These are synthetic public records. Do not use private documents in this public test flow.
Add knowledge from anywhere
Drop a UTF-8 file, fetch a public URL, or paste an IPFS CID. ECAI inspects the source first, shows exactly what it detected, then hands it to the durable searchable index queue.
Choose a source, then detect its structure before indexing.
ipfs_cid → searchable_disk job runs. Use Knowledge & privacy instead for confidential corpora.Wikimedia presets
Execution history
Open any job for progress, checkpoints, artifacts and live events. Tracking starts automatically and stops when you close it or switch jobs.
Mint preferences & review requestsFor completed index artifacts
Set account-wide review preferences here. Choose copies and metadata using the Mint button on a job. These settings and review requests do not submit a blockchain transaction.
Open to load your mint preferences.
Recent review requests
Open to load requests for this account.
Choose what to index
Pick one or more server-approved datasets, or configure an exact release and month window. Review the selection before it joins the index queue.
Wikimedia smoke sample
The local fixture includes Quantum mechanics, Classical mechanics, Physique quantique and a low-visibility Noise article. The smoke preset selects the top three and exercises download, ranking, indexing and finalization without an external Wikimedia download.
Select the smoke preset, add it to the queue below, then open View job. Search becomes available as records are indexed; previews are not final artifact proofs.
Pick Wikimedia projects
These presets are configured and validated by this ECAI node. Each selection becomes a separate durable indexing job.
Choose release and pageview months
Explore actual Cirrus release listings and suggested month windows. Month suggestions do not guarantee availability; Preview validates the underlying files before you queue.
Index doctor
No checks run yet.
Wikimedia index
No search yet.
Knowledge & privacy
Encode structured facts and access account-scoped corpora through authenticated REST operations.
Knowledge minting
The existing /ecai/ekef route requires a configured server-side signing keypair. Confirm any minting attempt before submitting.
No knowledge submitted.
Private corpus
The server derives the principal from DamageBDD authentication; do not submit owner IDs, credentials or keys in request JSON.
Responses are only held in this browser session.
Marketplace jobs
Experimental local chunk-job workflow. Disabled by default; set marketplace_enabled = true to enable. State is not durable, and marking a job paid does not transfer funds on-chain.
Published work
| ID | Status | Owner | Reward | |
|---|---|---|---|---|
| Sign in to view marketplace jobs. | ||||
Create chunk work
Claims & local status
Select a marketplace job to inspect or update it.
No job selected.
Ingestion & runtime
Inspect Yelp loader state and run server-side ingestion and publication steps deliberately.
Pipeline status
Status not loaded.
No chunk job selected.
Start chunking
No operation run.
WebSocket inspector
Not connected.
Code learning & repair
Inspect durable learning and repair activity, review verifier-validated Git diffs, and publish human-approved changes to a dedicated origin review branch. Nothing publishes automatically.
Code intelligence & repair workers
Code intelligence at a glance
Detailed node status · raw JSON
No administrator status loaded.
No repair status loaded.
RUNTIME POLICY · COLLABORATIONLearning, repositories & review gate
Changes are durable and take effect without restarting ECAI. In-flight learning is allowed to finish; new generations use the updated policy. Git credentials are never stored here.
Where ECAI learns
Fetch & fast-forward
Repository status not loaded.
Verification & publication
Reviewers & publishers
These lists can only narrow existing DamageBDD code-admin authority. Empty lists allow every authorized code admin.
Share verified learning evidence
Bundles are commit-pinned snapshots without raw source bodies. They are never auto-imported into live learning.
Recent repair attempts
Patch reviews
Only successful verifier results can enter the queue. A review pins a SHA-256 hash and base commit. Independent approval is required before publication. Publication re-runs verification, creates a new commit in an isolated worktree, and pushes without force to the configured review remote and branch prefix.
Patch & verification
Review decision
The publisher must be a different authorized publisher from every approving reviewer. Approval alone never pushes. Push must also be enabled and the base commit must still match the configured review repository HEAD.
No audit events selected.
Model Internals
Control the interfaces this node actually exposes. Keep deeper runtime hooks explicit, isolated and testable.
Wired here, or runtime required?
Capabilities have not been loaded.
Generation overrides
Node-wide policy, not an account preference. Blank numbers inherit existing node/request settings. Saves affect new requests; in-flight retries keep their pinned revision.
Request preview
Validate the current draft and inspect its overrides. No model is loaded, no prompt is sent and no spend is initiated.
No preview requested.
Saved policy evidence
Not loadedThis SHA-256 fingerprint identifies the saved configuration. It is not a digital signature or proof that generated statements are true.
Inference targets
This page cannot edit endpoints, add credentials, select an arbitrary model or load executable adapters.
Recent code inference
The latest 32 successful calls observed on this node. This view resets on service restart; policy metadata on repair attempts follows the existing durable repair store. No prompts, generated code or hidden activations are displayed.
Fund an index.
Set a satoshi budget. Your approved DamageBDD nodes build and independently verify the segments.
Source & budget
Your indexing pool
Only operator-allowlisted nodes can join this private cluster. Each node opts in and proves its Lightning identity. Joining does not open a channel or send sats.
Channel capacity has not been checked. Routed payments do not require a direct channel.
First-time node setup
Configure indexing_worker_nodes on the coordinator. On each peer, enable index_pool_participant_enabled, set index_pool_account to its node-admin account and allow the coordinator in index_pool_coordinators. Mount the same shared artifact root on every node. Keep Erlang distribution private.
Funded jobs
Market Maker
Coinstore DAMAGE/USDT and BTC/USDT · inventory-aware geometric quoting · restart-safe paper performance.
DAMAGE vs BTC movement
Samples recorded by the node since the market maker started. No synthetic price history.
Current bid and ask ladders
Bid · buy DAMAGE
| Price (USDT) | Qty DAMAGE | Notional |
|---|---|---|
| No active bids | ||
Ask · sell DAMAGE
| Price (USDT) | Qty DAMAGE | Notional |
|---|---|---|
| No active asks | ||
Configure quote engine
Edits are checked by the Erlang worker, persisted to DETS and applied to the next fresh quote cycle. They cannot enable real exchange execution or alter starting balances.
Engine state
Pausing removes paper quotes, but continues tracking market data and P&L. Resume is rejected when the drawdown limit is breached.
- Current halt reason
- —
- Feed age
- —
- Feed errors
- —
- Execution
- Disabled · paper only
Live trading is not configured
These controls alter only deterministic paper quotes. No Coinstore private order API, real balances, exchange keys or live order placement are exposed.
All operator changes require a managed DamageBDD node-admin account and explicit bearer authentication.
Paper equity vs BTC benchmark
The BTC benchmark assumes holding the starting equity entirely in BTC. Paper fills occur only when explicitly supplied to the engine; prices alone do not imply fills.
Recent paper fills
| When | Side | Qty | Price | Notional | Fee |
|---|---|---|---|---|---|
| No confirmed paper fills | |||||
Elliptic-curve audit chain
Each accepted state transition is chained with SHA-256 and mapped to a P-256 point. The point proves consistency with its canonical input; it does not validate external market feeds or predict prices.
- Retained event count
- —
- Last digest
- —
- Curve state point
- —
Recent state transitions
| When | Kind | Hash |
|---|---|---|
| No events recorded yet | ||
A character.
A point of view.
Your AI.
Discover portable personality packages for your LLM. Choose an edition, pay in sats, and collect its on-chain record.
Find your personality.
Review the character, licence and available editions.
Scan the invoice with your Lightning wallet.
Download the package after payment and mint confirmation.
Purchases & receipts
Sign in with your managed account to purchase a personality or resume a checkout.
Your edition includes the package and rights described in its licence. Model inference is separate. Personality behavior can differ between models; private conversations and memories are not included in a transfer.
Mathematical paths
Explore exact isogenies, trace rational points, and challenge every claim.
Sign in with a managed account to run the mathematical pathfinder.
A mathematician’s workbench
Browse the complete collection
Edit a problem or import your own path
Find a connection over Q
Reachable models
Local classes group the retained models by exact Q-isomorphism. They are not global class identifiers. Different model IDs or equal j-invariants alone do not decide isomorphism.
Objects and transformations
Select a curve or arrow to inspect it. Layout shows composition order, not mathematical distance. The path is not a global minimum-degree claim.
Follow a rational point
Fractions use bounded BigInt arithmetic, never floating-point rounding. A point experiment checks that point only; it is not an all-points isogeny proof. Live replay checks the full polynomial identities.
Select an object
Equations, exact coefficients and replay checks appear here.
Evidence, not a universal badge
Recent mathematical jobs
Jobs are volatile: up to 8 per account, 32 per node, one hour. A node restart clears them. Export artifacts to keep your evidence.
LLM evaluation
Run reviewed scenarios against selected, recorded model targets. Separate executable checks, human judgement and paper-trading results.
Managed email sign-in is required. Wallet-only sessions cannot run evaluations or change policy.
Start a comparison
Each target receives the same selected cases and generation limits. Gold checks and future market data stay outside the model request. The account settings below select targets and repetitions.
Run history
Run report
Pass rates use all planned trials, including errors and interruptions. A critical failure blocks the gate. Citation checks validate reference IDs, not semantic truth. Trading is paper replay only.
Target scorecards
Failure explorer
Recommendations and slice metrics
Selected trial: input, output and check evidence
Choose Inspect on a trial.
Paired regression comparison
Same suite and split required. Incomplete pairs remain visible.
Versioned evaluation datasets
Import cases, fixed references, executable checks and anchored human rubrics. Editing any gold label, reference, reviewer list or case creates a new content-addressed suite. Explicit reviewer IDs grant access to that suite and its trials.
Examples are synthetic smoke tests, not production benchmarks. Review sampling, annotation quality and leakage before using a suite as a release gate.
Coverage & annotation approval
Inspect a suite to see its source, frozen references and annotation guidelines.
Independent review queue
Only explicitly authorized reviewers can inspect or rate a shared suite. Trial views omit other reviewers’ votes and provider identity. Votes are immutable and disagreements remain indeterminate.
Choose a review item.
Account evaluation settings
These preferences affect this account on this node. Node allowlists and ceilings always take precedence. Credentials stay in the existing server-side model configuration.
Effective settings
Node-wide evaluation policy
The same versioned fields appear under DamageBDD → Node administration → ECAI settings → LLM evaluation. Defaults disable evaluation and remote inference. Changes are persisted by the existing node settings service.
Target IDs must refer to explicitly configured code_model_nodes or code_ollama_nodes entries. Remote targets require verified TLS. Limits cover this node; they are not a cross-node or currency budget.
Compute
Connect a cloud provider, set compute limits and reserve capacity for eligible index jobs.
Sign in to manage compute connections and reservation limits.
Loading cloud setup… If this message remains, reload the dashboard. A cached or missing dashboard asset may be preventing setup from initializing.
Choose a compute provider
Choose automatic AWS provisioning or delegate an existing environment manually. Only the dedicated automatic setup form accepts AWS credentials, for a temporary session; provider settings never store them.
- 1Choose provider
- 2Delegate access
- 3Verify identity
Connect once. Create your AWS environment.
ECAI generates the trust ID, network, private bucket, scoped IAM roles, queue and job definitions. Review the exact plan before approving AWS resource creation.
- 1 · Connect AWS
- 2 · Review plan
- 3 · Create & verify
1. Authorize this setup session
Use a setup-authorized IAM identity, never root keys. Credentials remain in a short-lived setup process, not your provider settings. Identity confirmation alone does not verify provisioning permissions.
2. Choose capacity and review
No resources are created during planning. The approved image comes from node configuration. Network: dedicated public subnets, no inbound worker access, outbound HTTPS, no NAT gateway.
Existing account policies stay unchanged. Storage, logs, networking and separately approved diagnostic jobs can incur AWS charges outside the indexing reservation ledger.
3. Review, create and verify
Resources & immutable deployment evidence
Optional paid verification job
Runs one synthetic S3 read/write job using the runtime role: 1 vCPU, 2 GiB, one attempt and a 300-second running timeout. Queue time is separate. AWS compute, IP and storage charges apply outside indexing reservations. Passing this test does not enable a production index dispatcher.
Not requested.
Remove environment · retain stored artifacts
Reconnect AWS setup credentials for this account and region first. Deletion removes the stack's compute, network, roles and logs. The private S3 bucket and its TLS-only policy are retained; stored data may continue to incur charges. Any active or uncertain verification job must be reconciled first.
Deployment events
Use existing AWS resources · manual setup
Connect AWS Batch
Start by preparing your account-specific external ID.
This node has not configured its coordinator IAM role. You can prepare a setup ID, but copying a safe trust policy requires compute_aws_coordinator_role_arn. Your AWS account cannot fix that node setting.
1 Prepare your AWS trust ID
Generate a stable ID before you create the role. It stays the same after you save or update the AWS connection.
2 Authorize ECAI in AWS
Create a cross-account IAM role with a custom trust policy restricted to your ECAI node's IAM principal and the external ID. Grant only required Batch/S3 permissions.
Copyable policy will be available once your node operator configures its coordinator role ARN.
3 Enter your existing AWS resources Role ARN, region, queue, job and S3 bucket
Connect Google Cloud Run Jobs
Use an existing Google Cloud project and workload identity.
1 Open your Google Cloud project
Select your project in Google Cloud. Create or choose an existing Cloud Run job with a pinned container image and a staging GCS bucket.
2 Delegate workload identity
Your node operator must configure workload identity federation or node-managed short-lived credentials. Set least-privilege Cloud Run and storage IAM permissions; do not upload a JSON key.
3 Enter your existing Google Cloud resources Project, region, job, bucket and service account
Spending limits
These are admission and reservation limits, not hard caps on your cloud provider's invoices. Configure provider billing alerts, job timeouts, quotas and spend limits independently. No billing reconciliation is active yet.
Remote index work
Only a node-approved shard_search Wikipedia/Yelp source job can be quoted. A reservation does not start a remote container; queue-dispatch integration is a separate step.
Reservations
API keys.
Create account-scoped bearer keys for Ollama-compatible clients. Keys can be rotated or revoked without changing your DamageBDD sign-in.
Create a key
The full secret is shown once. ECAI stores only a SHA-256 digest of the random secret plus non-secret metadata.
Quick configuration
——The quick-config route is GET /ecai/keys/ollama/quick-config. POSTing the same route creates a new key and returns ready-to-copy client configuration with the secret included once.
Copy this key now
This value cannot be retrieved again. It will be removed from this page automatically.
Ollama API keys
Use an existing key
Replace the placeholder with a key you just created. For custom hosts, use an Authorization bearer header. Keep credentials in environment variables rather than source code or browser bundles.
Loading…
Loading…
Loading…
Loading…
API explorer
Browse the routes exported by this ECAI node and make authenticated same-origin requests.
Choose an endpoint
Select a route on the left to inspect its parameters.
Select an endpoint to get started.
Account settings
Manage your session and find account-scoped preferences in one place.
Your session
Sign in to manage your account settings.
- Account
- Guest access
- Sign-in method
- Not signed in
- Account address
- Sign in to view your account
Switching accounts replaces this browser session; it does not link accounts. Never enter a wallet seed phrase or private key here.
Account preferences
Managed email sign-in is required for these settings.
Node settings
Find node-wide controls and diagnostics. Account preferences and cloud budgets remain separate.
About this node
- Node version
- Not reported
Node-wide changes require a server-authorized administrator.
Administrator controls
This session does not have administrator access. Sign in with an authorized account to see available node controls.
Open the existing runtime settings in Code learning & repair.
Administrator links do not grant permissions. The destination and server recheck access before any change.