Three research documents for Phase 2 planning:
- credential-provider.md: Outbound auth (CredentialProvider trait, CredentialSet enum),
account model as storage-layer concern (Identity.id as account UUID), SecretStoreCredentialProvider,
ManagedCredentialProvider, self-hosted service auth analysis (rustfs S3/OIDC, gitea OAuth2),
implementation phases A-D.
- interface-model.md: StreamInterface vs MessageInterface trait design, HTTP interface
as axum handler, DNS as MessageInterface, unified auth across all interfaces
(AuthToken + API keys via resolve_from_token), removal of TransportKind::Dns.
- tls-transport.md: Unified multi-interface architecture on port 443. Byte-peek protocol
detection (existing stealth mode) routes SSH vs axum. Axum multiplexes REST, WebSocket,
SSE, gRPC. QUIC/UDP with ALPN routing for WebTransport and iroh P2P. Single AuthToken
mechanism for all non-SSH interfaces. Four primitive operations (call/batch/schema/subscribe)
map to HTTP, MCP, and DNS.
Add doc comments and TODO markers to SshSession::recv() and send()
explicitly marking them as Phase 1 stubs. Notes that call protocol
event bridging from SSH channels is planned for Phase 2/3.
New Phase 1 modules should follow the existing pattern of referencing
ADR numbers in module-level doc comments for discoverability, matching
the style in transport/mod.rs.
ForwardingAction, TargetPattern, ForwardingRule, OperationType,
InterfaceConfig, InterfaceKind, DynamicConfig, and CallError are all
likely to gain variants/fields in future phases. Marking them
#[non_exhaustive] now prevents downstream breakage when new
variants/fields are added. Added constructor methods for types that
are constructed from other crates.
parse_proxy_config was using expect()/unwrap()/panic!() which would
crash the process on malformed proxy config strings instead of
returning a descriptive error. Now returns ConfigError::ProxyConfigInvalid
with the specific issue (bad scheme, bad address). Added tests for
invalid scheme, invalid address, and end-to-end from_serve_options.
NapiServerHandler was bypassing IdentityProvider, calling
config.auth.authenticate_publickey() directly, which meant no Identity
was stored on the session and per-identity forwarding rules could not
match. It also skipped ForwardingPolicy::check() entirely, defeating
forwarding access control for NAPI-served tunnels. Both are now
consistent with ServerHandler and SshHandler behavior.
Review found Phase 1 approved with minor issues. Created cleanup tasks:
- napi-identity-provider-wiring (bug: NAPI bypasses IdentityProvider)
- panic-free-static-config (code smell: panic/unwrap in production path)
- non-exhaustive-public-api (future-proofing public API types)
- adr-doc-comments (doc convention: ADR references in new modules)
- ssh-session-recv-stub-doc (documentation: mark stubs as planned)
- SshInterface implements Interface trait with accept() method
- SshSession implements InterfaceSession trait (stub for call protocol events)
- RawFramingInterface is type-only stub (Phase 4+ for DNS, WebTransport)
- TransportKind consolidated into transport module with Display, PartialEq, Eq
- ListenerConfig gains interface_kind field for (Transport, Interface) pairs
- SshInterface wraps existing russh handler logic (SshHandler)
- Auth delegation through IdentityProvider (not embedded in SshInterface)
- Channel routing through session to Layer 3 (forwarding policy)
- Server accept loop uses (Transport, Interface) pairs
Per ADR-026: SSH is Layer 2, not Layer 1. This is the highest-risk Phase 1
task, implementing the Interface trait to separate transport from interface.
- Add reloadAuth(), reloadForwarding(), reloadAll() methods to AlknetServer
- Add NAPI type definitions: AuthConfigNapi, ForwardingPolicyConfig, ForwardingRuleConfig
- Refactor NapiServerHandler to use ArcSwap<DynamicConfig> for atomic config swaps
- Add ConfigReloadHandle::dynamic_arc() accessor for sharing ArcSwap between NAPI and accept loop
- Add ipnetwork dependency to alknet-napi for TargetPattern CIDR parsing
- Add builder functions for AuthPolicy and ForwardingPolicy from NAPI config types
- All swaps are atomic via ArcSwap per ADR-030
Add Layer 2 interface abstraction per ADR-026:
- Interface trait with accept() and associated Session type
- InterfaceConfig enum with Ssh and RawFraming variants
- SshInterfaceConfig with auth, forwarding, host_key fields
- RawFramingConfig (minimal, no SSH-specific config)
- InterfaceSession trait with recv()/send() producing InterfaceEvent frames
- InterfaceEvent wraps EventEnvelope with optional Identity
- Resolves OQ-IF-01: every session produces EventEnvelope frames
via InterfaceSession, making Layer 3 interface-agnostic
- Valid (Transport, Interface) pair enumeration with
TransportKindBase and is_valid_pair validation function
- Module re-exported from lib.rs
- Change ServerHandler to hold Arc<dyn IdentityProvider> instead of Box<dyn IdentityProvider>
- Refactor Server::new() to use StaticConfig::from_serve_options() producing (StaticConfig, DynamicConfig)
- Remove duplicate parse_proxy_config from serve.rs (now in static_config.rs)
- Add with_identity_provider() accepting Arc<dyn IdentityProvider>
- Add integration tests for DynamicConfig reload and ForwardingPolicy deny
- Add test for custom IdentityProvider injection via with_identity_provider
- Move parse_proxy_config tests to static_config.rs module
Add local dispatch for OperationEnv with invoke() method, EventEnvelope
wire format struct, 4-byte BE length-prefixed frame encoding/decoding,
PendingRequestMap for call/subscribe correlation, call protocol event type
constants, and default /services/list and /services/schema operations.
Add ForwardingPolicy, ForwardingAction, ForwardingRule, and TargetPattern
types in config/forwarding.rs. Implement policy evaluation with first-match
wins semantics, principal and transport matching, CIDR and glob patterns.
Modify ServerHandler to check ForwardingPolicy before proxying in
channel_open_direct_tcpip. Reserved alknet-* destinations bypass policy.
Preserve existing behavior with default allow_all() policy.
Add AuthProtocol enum (VerifyPubkey, VerifyToken, ReloadKeys, CheckAccess),
AuthResult enum (Ok(Identity), Denied(String)), and AuthServiceImpl
wrapping ConfigIdentityProvider via ArcSwap<DynamicConfig>. All gated
behind the irpc feature flag per ADR-028.
- Add ListenerConfig struct with transport_kind, listen_addr, per-transport config
- Add Dns and WebTransport variants to TransportKind (tags only, no behavior)
- Add .listeners() builder method to ServeOptions for multi-listener config
- Keep .transport_mode() backwards compatible (creates single-element listeners vec)
- Update Server::run() to use listeners from Server struct (first listener)
- Add Server::listeners() accessor for multi-transport listener configs
- Update StaticConfig to support listeners field, converted from ServeOptions
- All listeners share Arc<ArcSwap<DynamicConfig>>, ConnectionRateLimiter, and IdentityProvider
- Graceful shutdown terminates accept loop via existing shutdown signal
- TOML [[listeners]] array-of-tables syntax supported via ListenerConfig in StaticConfig
- Add comprehensive tests for ListenerConfig, multi-listener ServeOptions, Server creation
Add Identity struct with id/scopes/resources fields and IdentityProvider
trait with resolve_from_fingerprint/resolve_from_token methods. Implement
ConfigIdentityProvider reading from ArcSwap<DynamicConfig.auth> for
fingerprint-based key lookups. Delegate ServerHandler::auth_publickey()
through IdentityProvider instead of direct AuthPolicy access. Store
authenticated Identity in the handler for use by ForwardingPolicy.
ConfigServiceImpl wraps ArcSwap<DynamicConfig> providing forwarding_policy(),
rate_limits(), and reload() methods for direct use (always available).
ConfigProtocol enum (GetForwardingPolicy, GetRateLimits, ReloadForwarding,
ReloadRateLimits) is gated behind the irpc feature flag per ADR-030.
Split alknet-core configuration into StaticConfig (immutable after startup)
and DynamicConfig (hot-reloadable at runtime via ArcSwap).
- Add StaticConfig struct in config/static_config.rs with all fields per ADR-030
- Add DynamicConfig struct with AuthPolicy, ForwardingPolicy, RateLimitConfig
- Add ForwardingPolicy with allow_all()/deny_all() defaults (ADR-031)
- Add ConfigReloadHandle with reload() method for runtime config updates
- Replace Arc<ServerAuthConfig> with Arc<ArcSwap<DynamicConfig>> in ServerHandler
- Add config_reload_handle() to Server for obtaining reload handles
- Add AuthPolicy with authenticate_publickey/authenticate_certificate methods
- All existing tests pass with the new config structure
- Default DynamicConfig produces identical behavior to current code
Phase 1 of the integration plan modifies alknet-core to support the
architectural changes from Phase 0 ADRs and specs. Decomposed into
dependency-ordered tasks across config split, identity, forwarding
policy, OperationEnv, interface abstraction, and NAPI reload API.
Critical path: config-split → identity → forwarding → wire-into-handler
→ interface-trait → ssh-interface-extraction → review.
Two highest-risk tasks (interface-trait-definition, ssh-interface-extraction)
are split from §1.8 per the integration plan's note that it may need
sub-phases. OperationEnv is split into types and runtime per Phase 1
local-dispatch-only constraint.
Update four existing specs (overview, server, napi-and-pubsub, call-protocol) to
reflect Phase 0 decisions: three-layer model, IdentityProvider, ForwardingPolicy,
OperationEnv, static/dynamic config split. Review all 9 Phase 0a ADRs (026-034)
for consistency. Fix 4 critical issues from architecture review: missing OQ-SVC-05
in open-questions.md, deprecated hub terminology, undefined AuthService and noq
terms. Replace inline OQ text with cross-references per format rules. Add
ConfigServiceImpl definition to configuration.md. Port absolute workspace paths
to project-relative links by copying referenced docs (feasibility, certbot,
fail2ban, event_source_types) into docs/research/.
The remaining task descriptions implied that downstream concerns
(StorageIdentityProvider, irpc service layer, agent services, multi-node
deployment) already exist. Updated to clearly distinguish:
- spec-update-server: Phase 1 ships ConfigIdentityProvider, not irpc auth
- spec-update-call-protocol: Phase 1 is local dispatch only; irpc and remote
dispatch are contracted for later. Agent services are downstream concerns.
- spec-update-overview: Note which crates exist now vs which are Phase 2+ contracts
- review-spec-foundation: Add phase boundary check to acceptance criteria
The architecture specs were implying that StorageIdentityProvider, irpc
service implementations, and application services (agent, Docker, etc.)
already exist. This commit makes the phasing explicit:
- services.md: deployment topology now clearly labels 'Current (Phase 1)'
vs 'Future (Phase 2+)', notes that application services are downstream
- identity.md: StorageIdentityProvider labeled 'Future — Phase 2+',
clarifying alknet-storage doesn't exist yet
- storage.md: adds phase note that the crate hasn't been built yet,
StorageIdentityProvider is a future impl
- ADR-028: ConfigAuthService is Phase 1 path, StorageAuthService is
Phase 2+ contract
- call-protocol.md: Agent Service Pattern section explicitly framed as
a downstream application concern, not a core requirement
Add 10 new tasks under tasks/architecture/ for Phase 0a (ADR writing):
- 9 ADR tasks (026-034) with dependency-ordered structure
- 1 review checkpoint task before Phase 0b spec writing
ADR dependency graph (3 generations):
Gen 1 (parallel): 026, 029, 030, 031, 032, 034
Gen 2 (depends on 029): 027, 028
Gen 3 (depends on 027+028): 033
Gen 4: review checkpoint
Also mark all 34 prior implementation tasks as completed — they
were finished but still showing as pending in the taskgraph.
- Replace hub/spoke with head/worker terminology in call-protocol.md,
auth.md, open-questions.md, napi-and-pubsub.md
- Update operation paths from /{spoke}/{service}/{op} to
/{node}/{service}/{op} throughout call-protocol.md
- Unify Identity struct: auth.md already had {id, scopes, resources},
add note clarifying this is canonical (vs research/services.md which
used {node_id, fingerprint, scopes})
- Update integration-plan.md inconsistencies section to track what's
been fixed (hub/spoke, identity model) and expand service naming
to include external services
- Update call-protocol.md last_updated date
ADRs are intentionally left unchanged as historical records.
Organizes findings from the research phase (core, services, configuration,
storage, flow) into an actionable phased plan covering:
- Transport/Interface/Protocol three-layer model
- OperationEnv as universal composition mechanism (not replaced by irpc)
- Phase 0: Architecture foundation (9 ADRs, ~10 spec docs)
- Phase 1: Core modifications (config split, identity, forwarding, auth,
OperationEnv, interface abstraction)
- Phase 2: External crates (alknet-secret, alknet-storage, alknet-flowgraph)
- Phase 3: Integration and wiring
- Phase 4: Advanced features (DNS, WebTransport, app services)
Key clarifications: irpc services are one dispatch backend for OperationEnv,
not a replacement for it. DNS control channel is a (DNS transport, raw framing
interface) pair, not SSH-over-DNS. Call protocol and irpc operate at different
scope boundaries within Layer 3.
Document the OperationContext (request_id, identity, metadata, env, trusted),
OperationEnv (namespaced callables for handler composition), ResponseEnvelope
pattern, and how MCP/OpenAPI adapters map to the irpc service model.
- Replace hub/spoke terminology with head/worker throughout all research docs
- Add irpc service layer architecture (AuthProtocol, SecretProtocol,
ConfigProtocol, StorageProtocol)
- Add BIP39/SLIP-0010 HD key derivation for secrets management
- Add event boundary discipline (domain events vs integration events)
- Add application services layer (Docker, Node, Wallet, Proxy, Compute)
- New docs/research/services.md defining irpc service protocols
- Update core.md with service layer section and head/worker model
- Update configuration.md to delegate auth to AuthService (irpc)
- Update storage.md with secrets/key derivation and event boundaries
- Update flow.md with event boundary decision and cross-references
Unified authentication (ADR-023): SSH and WebTransport auth share the same
Ed25519 key material. Token auth uses signed timestamps verified against the
same authorized_keys set. IdentityProvider trait decouples core from identity
storage.
Bidirectional call protocol (ADR-024): Generalizes control channel (ADR-018)
to support hub→spoke and spoke→hub calls. Operation paths use /{spoke}/{service}/{op}
format for three-level routing. EventEnvelope wire format, five call events,
PendingRequestMap for correlation.
Handler/spec separation (ADR-025): Downstream consumers register operations
without modifying core. OperationRegistry maps paths to specs + handlers.
Service discovery via /services/list and /services/schema.
Resolves OQ-17 (transport-aware auth), OQ-21 (spoke routing), OQ-CFG-04 and
OQ-CFG-06 (WebTransport auth and transport-aware auth layer). Adds OQ-18
through OQ-22 for remaining open questions.
Enable running wraith alongside iroh-blobs, iroh-gossip, and iroh-docs
on the same QUIC endpoint (one connection per peer, multiplexed by ALPN).
- IrohTransport::from_endpoint(node_id, endpoint) for client-side shared endpoint
- IrohAcceptor::from_endpoint(endpoint) for server-side shared endpoint
- Export ALPN constant as IROH_ALPN for Router registration
- Add owned() method to track whether the endpoint was created internally
- Existing new()/bind() constructors unchanged (backwards compatible)
- Add tests for from_endpoint constructors and shared endpoint connectivity
- Restrict auth methods to PUBLICKEY only (no none, password, hostbased,
or keyboard-interactive advertised during negotiation)
- Log all denied channel types (session, x11, forwarded-tcpip) and
dangerous request types (exec, shell, subsystem, pty, env, x11, agent)
- Explicitly reject all dangerous channel request handlers (exec, shell,
subsystem, pty, env, x11, agent forwarding) with channel_failure
responses instead of russh's default silent Ok(()) which leaves clients
hanging and is a footgun if session channels are ever allowed
- Explicitly reject tcpip_forward, streamlocal_forward with logged warnings
- Log signal requests at debug level (harmless, no response needed)
- Override handlers in both core ServerHandler and NapiServerHandler
- Add tracing dependency to wraith-napi for security event logging
- Set preferred algorithms explicitly (russh::Preferred::DEFAULT which
uses only modern KEX/cipher/MAC algorithms)
Add top-level README.md with alpha status warning, quick start guide,
architecture overview, feature flags, transport modes, auth docs, and
Node.js API examples.
Add dual LICENSE-MIT and LICENSE-APACHE files.
Add comprehensive crate-level and module-level rustdoc to all three
crates (wraith-core, wraith, wraith-napi) and all public modules
(transport, client, server, auth, socks5, error). Add doc comments to
key public types (Transport, TransportAcceptor, ConnectOptions,
ClientSession, Server, ServeOptions, KeySource, ServerAuthConfig, etc).
Update Cargo.toml files with workspace-level package metadata
(version, edition, license, repository) and crate descriptions.
- handler.channel_open_direct_tcpip now proxies non-wraith channels via
connect_outbound+proxy_channel instead of dropping them
- ClientSession.run() spawns reconnect monitor that detects handle closure,
reconnects with exponential backoff (1s/2s/4s/8s/16s/30s cap),
and re-registers remote port forwards
- Remove server-side logging of tunnel destinations (ADR-006 compliance)
- Remove debug-level logging of proxy targets in channel_proxy
Expose NAPI serve() per ADR-016. WraithServer provides close() and
onConnection(callback) for receiving SSH channel streams from
incoming connections. Each connection produces a WraithServerStream
(Duplex-like read/write/close) with ConnectionInfo (remoteAddr,
transportKind). Supports TCP transport with optional authorizedKeys
and certAuthority auth. TLS and iroh transports return helpful errors
indicating future support.
Add stealth mode detection that peeks at the first bytes after TLS handshake
to determine SSH vs HTTP protocol. SSH connections proceed to russh handler;
non-SSH connections receive a fake nginx 404 response, making the server
indistinguishable from an ordinary HTTPS site to scanners and DPI systems.
- ProtocolDetection enum (Ssh, Http) for protocol classification
- detect_protocol() uses BufReader::fill_buf() to peek without consuming bytes
- send_fake_nginx_404() writes HTTP/1.1 404 + Server: nginx headers
- validate_stealth_config() enforces TLS transport requirement for stealth
- 17 unit tests covering SSH banner, HTTP, random data, and edge cases
- Add WraithConnectOptions struct with napi fields: server, peer, transport,
identity (string path or Buffer), tlsServerName, insecure, irohRelay, proxy
- Add WraithStream napi class wrapping SSH channel read/write halves via
ChannelStream::into_stream() + tokio::io::split()
- Implement connect() async function: transport creation (tcp, tls), SSH client
connection, authenticate, open direct_tcpip channel, return WraithStream
- Identity field accepts file path (string) or in-memory key data (Buffer)
- All Rust errors marshalled to JavaScript exceptions with descriptive messages
- Add ForwardError enum to wraith-core (required by forward.rs)
- Enable tls, iroh features on wraith-core dependency
- 7 unit tests for key source resolution and address parsing
Adds client/connect.rs with ConnectOptions (programmatic API per ADR-011),
ClientSession::new() for SSH session establishment, ClientSession::run()
for SOCKS5 + port forwards + shutdown, and graceful shutdown via
SIGTERM/SIGINT with SSH disconnect and 2s drain timeout.
Add ConnectionRateLimiter (HashMap<IpAddr, usize>) and AuthAttemptLimiter
with check/on_connect/on_disconnect and check/on_failure methods.
Integrate into ServerHandler with structured tracing::info! logging for
auth attempts, connection opened/closed events. No logging of tunnel
destinations per ADR-006. Also add ForwardError type and fix type
annotation in forward.rs to unblock compilation.
Add AcmeCertProvider with domain-based and IP-based modes using rustls-acme.
AcmeTlsAcceptor::bind_acme() and TlsAcceptor::bind_acme() provide ACME-integrated
TLS acceptance with automatic cert renewal via background tokio task.
Feature-gated behind 'acme' (implies 'tls'). Unit tests for config construction;
integration test for LE staging marked #[ignore].
Convert socks5.rs to directory module with protocol parsing and server
implementation. Socks5Server binds to configurable address (default
127.0.0.1:1080), handles SOCKS5 handshake (no-auth), parses IPv4/IPv6/domain
addresses, and proxies bidirectionally via SSH direct_tcpip channels.
Domain names sent unresolved (SOCKS5h) to prevent DNS leaks (ADR-006).
No logging of request targets per privacy requirements.
Add iroh QUIC P2P transport using tokio::io::join for stream duplexing
per ADR-003. Default relay is n0's https://relay.iroh.network/ (ADR-009).
Proxy URL passed to Endpoint::builder (ADR-010). Integration test marked
#[ignore] for CI since it requires iroh relay connectivity.
Address 5 critical and 7 warning issues from architecture review:
- Fix duplicate sentence in napi-and-pubsub.md server side section
- Add wraith- namespace reservation to server.md constraints (ADR-018)
- Document stealth mode TLS-only requirement in server.md
- Create ADR-019 for --proxy dual semantics (client vs server)
- Clarify NAPI connect() vs CLI wraith connect distinction
- Add SOCKS5h default as privacy design decision in client.md
- Expand reconnection section (always-on, re-register port forwards)
- Add graceful shutdown sections to client.md and server.md
- Specify OpenSSH key format for path-or-buffer inputs across all docs
- Resolve pubsub alternative approach ambiguity (ADR-018 is primary)
- Replace server.md handler impl block with behavioral description
- Standardize iroh endpoint ID terminology (base58-encoded)
- Remove iroh API implementation details from transport.md/server.md
- Add error handling pattern as cross-cutting concern in overview.md
- Update all document statuses from draft to reviewed
glm-5.1