Fewer billable retries.
Fewer repeated case files.
A reason when T.38 fails.

Your receiving contract should not decide how every outbound fax travels.

Keep inbound and outbound providers independent. Add authorized outbound routes, set a preference for a destination, and retain the provider and configuration bound to every attempt.

Cloud adapters, the built-in Asterisk engine and the optional SSL Fax engine have different capabilities. An extra route is not another instance of the built-in SIP trunk.

A cheap route that repeatedly fails adds attempts, charges and delay.

Rank routes using known rates and recent delivery history for the destination. Record why a route was selected and use confirmed failure to move to another route.

Once two reliable routes have each delivered three faxes to a number, Faxbot sends by the one that costs less per delivered fax. Until then, rate cards and the reliability threshold decide, and an unknown outcome never triggers a resend.

A cloud fax page fee can sit on top of telephony you already pay for.

Run the fax engine against a carrier SIP trunk or connect it to Avaya IP Office or Aura on the local network. Separate fax processing from the carrier’s call charge.

T.38 and G.711 depend on the complete media path. Avaya, BT, Gamma and Telstra presets have local contract tests; real-system sign-off remains separate.

SIP registration works, then T.38 stalls behind a Docker network, port-changing NAT or a carrier that declines the re-INVITE.

Check network reachability, port mapping and the fax engine. Send the first T.38 packets so a port-preserving router opens the path. For Telnyx, check that each trunk number has the T.38 gateway turned on and turn it on from the trunk page. Show the call’s failure reason, select audio for new calls when the network cannot carry T.38, and restore T.38 when the path recovers.

Network reachability and carrier negotiation are separate checks. Changing the mode does not retransmit the failed fax. Sending over Telnyx T.38 is proven live; receiving over T.38 is newly live and still being hardened.

A conventional fax call spends most of its billed time pushing pages at modem speed, even when the far machine could take them faster.

Run the optional HylaFAX+ engine beside Asterisk on the same trunk. It offers SSL Fax inside the fax call. Every engine call records the speed, compression, error correction and seconds per page it actually negotiated; each fax says how its call went, and the trunk page groups the calls so you can see what costs time.

Faxbot only measures for now; nothing changes how it sends. The built-in Asterisk engine reports only the last page’s speed and resolution. SSL Fax works only when the far machine offers it, and so far it has been tested only in a lab loopback; its pages travel encrypted, but the far end is not verified. The engine itself has sent and received live faxes over Telnyx on audio and T.38 calls.

Different locations, vendors and accounts need distinct media paths, capacity and access rules.

Extend route configuration into independent account and engine instances, with explicit inbound mappings, authorized outbound choices, capacity and policy scopes across the fleet.

The current built-in Asterisk connection and trunk are scalar. Multi-trunk composition and location-based route authorization are tailored to your sites, account boundaries and permitted paths.

T.38 relay, Internet Aware Fax and SSL Fax are often presented as interchangeable switches.

Test T.38 Internet Aware Fax on compatible endpoints before offering it, next to the SSL Fax engine, using the speed, compression and error-correction profile Faxbot already records for each destination.

The phone leg stays connected. T.38 support alone does not establish Internet Aware Fax or SSL Fax support.

Failed attempts, delayed charges and an inbound hand-over failure can leave the carrier bill out of step with the console.

Reconcile SignalWire fax charges and Telnyx trunk calls. Keep estimates, reported charges and settled amounts separate. Retain corrections and unmatched carrier calls; count charged failures.

An ambiguous call stays unmatched. Reconciliation for other trunk carriers is tailored to their billing interfaces and your account. Delivery status does not change because a charge arrived.

Per-call minimums and billing increments can erase a transmission-time improvement. Included usage can make a different route cheaper at the margin.

Estimate in the rate card’s actual unit, apply its minimum and increment to each attempt, and represent flat monthly plans separately from incremental fax charges.

A monthly fee remains payable when a fax adds no charge. Advertised rates are editable observations, not your contract or a savings guarantee.

Repeated short documents to the same destination keep paying connection overhead.

With recipient agreement, hold short faxes for a bounded interval and send them in one SIP call. Set the page cap and sender-mixing policy; Send now bypasses the wait.

Separators count as pages. Each document keeps its own confirmed-page result and charge share. A partial or uncertain shared call is not automatically retransmitted.

Rate-card order misses negotiated speed, long retries and route-specific paid failures.

For each number and route, add up what every attempt in the last 30 days really cost, failed and uncertain ones included, and divide by the faxes delivered. Recipients and Costs → Recommendations show the figure and offer the cheaper route.

The automatic route choice has been tested on simulated traffic, not yet on live traffic. Faxes sent through a monthly plan show as included rather than free. Noticing when a path changes is still planned.

A plan sized for heavy use keeps billing the same fee when traffic is light.

For each monthly plan, compare the fee per fax it actually carried with what the same faxes would cost on a reliable metered route, and say what else the plan provides, such as its fax number.

Advice only: Faxbot never cancels a plan. A number people still fax has to be moved to the other carrier first. Faxes to your own numbers count as tests.

Test and internal faxes to the installation’s own numbers pay for two call legs and can still fail on the carrier path.

When a fax is addressed to one of this installation’s own receiving numbers, Faxbot puts it straight into Received, once, with its page count and checksum. Test faxes that must prove the carrier path can still ask for a real call.

It applies only to numbers that receive faxes into this installation. Scheduling per destination is still planned.

A fleet pays fixed channels for quiet DIDs, while busy numbers still encounter contention.

Faxbot replays your received calls to find which numbers could share lines and which should stay billed by the minute, and lists quiet numbers with their monthly rental, using Telnyx’s published prices.

This is advice only. Faxbot never changes your carrier account, and it never advises a choice that would have given a caller a busy signal. A full pool turns callers away; it does not overflow to metered lines. It needs 60 days of call history and has been tested only with simulated traffic so far.

Retries pile onto a busy destination; an uncertain call can still occupy a channel after a worker stops waiting.

Faxbot dials a number only when it has room: one call at a time to each fax number by default, no more calls than the trunk’s fax lines, and no faster than the carrier’s published call rate. A call whose result is unknown keeps its place until it is settled, urgent faxes go first, and senders take turns.

A fax that has to wait says why and never fails because of it. The limits come from published carrier figures and your settings; so far they have been tested with simulated traffic and one live check against a busy fax number.

Every update repeats the growing packet, including documents the recipient already accepted.

For recipients that accept references, preview an index plus new or revised originals against the case’s accepted-document history.

The administrator records recipient agreement. Fax transport acceptance is not proof of chart filing or retention. Reuse stays specific to the case and recipient.

Rasterizing a signed or searchable PDF loses useful content and pays for a telephone transfer.

Enrolled Faxbot partners exchange encrypted original PDFs and signed durable-acceptance receipts using the recipient’s usual fax number. Ordinary fax remains the route for other recipients.

Both installations must enroll and verify the number and recipient binding. A lost receipt requires reconciliation before fax fallback.

A provider says received, but the PDF download or engine hand-over fails. Restarting can make the record disappear or duplicate it.

Persist acquisition state and resume fetching the real document. Preserve provider/account identity and its digest. Recover trunk images that were not handed over to Received.

The item waits until acquisition is checked. A notification never substitutes for document bytes.

The server accepted the request, but the console, SDK or assistant lost its response.

Keep a stable send-operation ID and recover the accepted job through the console, SDKs and MCP tools. Preserve transport uncertainty independently of client request recovery.

An intentional new send needs a new operation identity. Smaller one-bit PDFs preserve source image geometry; they do not promise fewer fax pages.

Large packets hit provider caps, interrupted calls and ambiguous partial delivery.

Distinguish sender continuation from recipient-verified repair. For enrolled peers, transfer missing components and commit a whole document only after checking order, length and hashes.

Atomic receiver assembly needs a compatible recipient and integration tailored to its intake. A transport page count alone cannot prove the recipient has a complete packet.

Staff rebuild giant packets and duplicate cover sheets without a dependable recipient checklist.

Compile unchanged source documents against an approved, versioned checklist. Show missing items and provenance. Add send-once internal distribution only where the recipient organization agrees.

Checklist compilation and internal distribution are tailored to the recipient’s approved requirements. A model must not summarize away required evidence or infer consent from a shared brand.

An intake workflow requires a fax arrival even when the enrolled recipient can receive the PDF digitally.

Send the encrypted original directly and one opaque notice page through fax. A recipient connector retrieves and checks the original for the existing intake workflow.

Notice hydration requires a connector tailored to the receiving workflow. The notice, original acceptance and downstream import need separate receipts; the notice does not prove the complete file was faxed.

Repeat partners exchange the same form templates and unchanged pages over and over.

Use exact content fingerprints, recipient inventory and a full-repair path for missing bases. Investigate lossless patches only where both ends can reconstruct and verify the original.

Component reuse requires integration with the recipient’s inventory; fax-carried patches remain experimental. Digital byte savings do not remove another fax call after direct delivery has already removed it.

Adding another integration or assigning an owner must not silently widen document access.

Use server-enforced users, groups, roles, mailbox grants and scoped API keys. Console, CLI, SDK and MCP calls use the caller’s permissions. Assignment requires existing access.

One installation serves one organization. Policy scopes do not establish tenant isolation or arbitrary per-location trunk routing.

Documents reach an inbox, then sit unowned while staff assume another team is responsible.

Deliver acquired PDFs through the configured SMTP intake connector. Assign an owner, set an acknowledgement target and backup, and export the item’s evidence. Import PDFs from another system through the generic import API.

Transport, SMTP acceptance, owner acknowledgement and completion are distinct events. A generic import is not an EHR connector.

Matching, splitting multi-patient documents, chart import and staff notification each fail differently. Adding a second portal can leave the actual work untouched.

Build connectors around the installed product’s enabled interfaces, with reviewed matching, selective redaction, idempotent writes and observed filing and task receipts. Add workforce identity and processing controls.

EHR/case connectors and controlled OCR/AI are scoped for the installed version, enabled interfaces and processing policy. Existing native features must be checked first. Generic import does not claim Epic, Intergy or other vendor integration.

Sites differ in identity, approvals, retention, purpose, recipients and processing constraints; staff should not reinvent policy on every send.

Use reviewed setup proposals built from the actual stack and versioned templates. Preview an inactive configuration, resolve required facts, approve the exact action, then apply with revision checks.

Guided proposals, approval policy, holds and jurisdiction templates are tailored to the organization’s requirements. Suggestions must not invent an API, grant access, assume legal defaults or silently move processing to the cloud.