@crvouga/mockingbird-service-hermes 1.0.0 → 1.1.0

package/API_EVIDENCE.md

@@ -1,6 +1,6 @@
 # Hermes peer-run API evidence
 
-Research for the public tracked-run mock; no Hermes runtime, inference, provider
+Research for the public tracked-run emulator; no Hermes runtime, inference, provider
 credentials, or Kanban dispatcher was executed. Source inspection is not observed
 parity. The package remains unimplemented until the later implementation stories.
 
@@ -69,7 +69,7 @@ A `_request_agent_overrides`. Explicit truthy history must be an array of object
 with `role` and `content`; those values are coerced to strings. Explicit history
 wins over previous-response history. A missing previous-response record is ignored
 by this pinned run handler, unlike the separate Responses API. Transcript loading,
-model routing and execution are not inference requirements for the mock; retain
+model routing and execution are not inference requirements for the emulator; retain
 synthetic metadata and script public observations without executing them.
 
 A new run has an opaque `run_` plus UUID-hex ID. The POST response is exactly
@@ -94,7 +94,7 @@ adds `error`; approval state can add `approval`; completion can include
 
 Public states observed in source are `queued`, `running`, `waiting_for_approval`,
 `stopping`, `completed`, `failed`, `cancelled`, and `interrupted`. Only the latter
-four are terminal for durable replay. The mock can script observations without
+four are terminal for durable replay. The emulator can script observations without
 implementing tools, approval resolution, SSE, steering, or agent execution.
 
 ## Errors and overload
@@ -147,7 +147,7 @@ For ordinary API calls the ownership/replay scope is SHA-256 of
 the expected configured API key or the literal unauthenticated-listener fallback.
 It is not derived from arbitrary supplied bearer text. Hosted-room requests use a
 separate verified-claims scope (room/install/authority epoch/member/target/profile);
-room token generation and validation are excluded. The mock must model isolation
+room token generation and validation are excluded. The emulator must model isolation
 using explicit synthetic profile/identity controls, without persisting credentials
 or changing consumer authorization. [R `_run_idempotency_scope`.]
 
@@ -187,7 +187,7 @@ returns409 as above. [R lines1353–1422.]
 When a stopped executor returns `interrupted: true`, status becomes `cancelled`.
 If completion wins and returns a normal final result, it becomes `completed`,
 even after a stop request. Structured execution failure remains `failed`.
-Stopping queued work before agent creation also ends as cancelled. The mock must
+Stopping queued work before agent creation also ends as cancelled. The emulator must
 script these distinct outcomes and preserve pending work between acceptance and
 settlement. [R lines705–719 and858–927; T
 `test_completion_wins_before_uncooperative_stop_is_acknowledged`.]
@@ -263,7 +263,7 @@ Resolved `/nousresearch/hermes-agent` again and queried submission/polling field
 and validation. Context7 still lacks `v2026.8.31`; results cite current-main run
 handlers and programmatic-integration documentation. Its newer terminal flags
 (`completed`, `partial`, `turn_exit_reason`) and shutdown wording are not imported
-into this pinned mock. Re-fetched R at immutable commit
+into this pinned emulator. Re-fetched R at immutable commit
 `29112bef099274229cadff79cdff7bf7b99c4b77`; SHA-256 remains
 `048ae843592d701ff47437bd8edd47cd64ca6c0fdf88a71045bdb6fb337fbc63`.
 Re-read admission, `_set_run_status`, completion/failure and GET branches.
@@ -271,7 +271,7 @@ Re-read admission, `_set_run_status`, completion/failure and GET branches.
 The implemented subset admits keyless ordinary submissions, preserves session/model
 metadata, and polls stored public observations. Python JSON truthiness is retained
 for input/history validation. Non-object roots and malformed final message elements
-return explicit mock-only 501 responses because their upstream error behavior is
+return explicit emulator-only 501 responses because their upstream error behavior is
 not verified. Hosted rooms, nonempty idempotency keys and invalid memory headers
 also return explicit 501 responses at this stage. Valid memory keys are echoed;
 no credentials are checked or used as namespace identity.
@@ -379,7 +379,7 @@ conflict is returned. Hosted-room retention and SSE buffer expiry are excluded.
 
 Tests exercise exact equality and one millisecond beyond, cache recovery from
 durable rows, pruning versus cached survivors, active and interrupted states,
-progress persistence, restart and Timeline. The mock folds elapsed sweep ticks
+progress persistence, restart and Timeline. The emulator folds elapsed sweep ticks
 into the next admission/observation and offers an explicit single-sweep control;
 it does not start a background worker. This source-backed model remains subject
 to the independent runtime comparison in US-022.

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog — @crvouga/mockingbird-service-hermes
 
+## 1.1.0 (2026-10-09)
+
+### Features
+
+- redesign emulator experience ([d6038c2](https://github.com/crvouga/mockingbird/commit/d6038c27eebbaa1751f02bc9305002a5b1304779))
+
 ## 1.0.0 (2026-10-07)
 
 ### ⚠️ Breaking changes

package/DISCOVERY.md

@@ -1,7 +1,7 @@
 # @crvouga/mockingbird-service-hermes discovery
 
 This is the installed-package index for coding agents and tooling. All relative links resolve
-inside `node_modules/@crvouga/mockingbird-service-hermes/`; no repository checkout is needed to discover the mock's
+inside `node_modules/@crvouga/mockingbird-service-hermes/`; no repository checkout is needed to discover the emulator's
 supported surface or documented behavior.
 
 ## Capability and behavior sources
@@ -47,7 +47,7 @@ designed for assertions and diagnosis by consuming test suites.
 ## Report a mismatch or missing capability
 
 Follow the [agent reporting contract](https://github.com/crvouga/mockingbird/blob/main/docs/REPORTING_ISSUES.md). Include package version,
-operation/command, a minimal redacted request, actual mock result, expected oracle result or vendor
+operation/command, a minimal redacted request, actual emulator result, expected oracle result or vendor
 documentation, and whether the mismatch appears in the matrix. Never include keys, tokens,
 customer data, prompts, PHI, card data, or unredacted recordings.
 

package/README.md

@@ -1,8 +1,8 @@
 # @crvouga/mockingbird-service-hermes
 
-> Familiar calls. Faithful echoes. Part of [Mockingbird](https://github.com/crvouga/mockingbird).
+> Local emulators. Real API contracts. Part of [Mockingbird](https://github.com/crvouga/mockingbird).
 
-Work-in-progress mock for the Hermes Agent public peer-run API pinned to
+Work-in-progress emulator for the Hermes Agent public peer-run API pinned to
 `v2026.8.31`. Submission, polling and stop work with explicit synthetic lifecycle observations.
 Events, approval and steer remain unsupported. No agent or inference runs.
 [API_EVIDENCE.md](API_EVIDENCE.md) records source-backed semantics and gaps;
@@ -50,10 +50,10 @@ The Node entry is separate from the portable Fetch entry.
 
 `POST /v1/runs` admits a run and `GET /v1/runs/{run_id}` polls it.
 `POST /v1/runs/{run_id}/stop` requests interruption. Events, approval and steer
-return a mock-only 501 envelope with `error.type` of
+return an emulator-only 501 envelope with `error.type` of
 `mockingbird_unsupported` and `error.code` of `operation_not_implemented`.
 Missing runs use the pinned `run_not_found` 404 envelope. Unknown paths return
-404. Mock-only errors do not claim real Hermes rejection behavior.
+404. Emulator-only errors do not claim real Hermes rejection behavior.
 
 - `GET /__admin/health` identifies the `hermes` runtime.
 - Select isolated namespaces with `x-mockingbird-namespace` or `/__admin/ns/<name>/…`.
@@ -128,13 +128,13 @@ The observation control accepts `status`: `queued`, `running`,
 for completed. Usage has three nonnegative integer counts: `input_tokens`,
 `output_tokens`, `total_tokens`. Completion defaults to empty output and zero counts.
 Terminal observations cannot be changed (409); invalid control payloads return 400
-without changing state. These are mock controls, not additional Hermes routes.
+without changing state. These are emulator controls, not additional Hermes routes.
 
 Timestamps are Unix seconds from the shared clock. Run records and their identity
 sequence use shared storage and Timeline; reset clears them. No timers, agent
 handles or prompts are stored. Polling result content stays out of journals.
 Malformed roots/final input elements, hosted rooms and invalid memory-scope headers
-currently return explicit mock-only 501 responses. See evidence for these limits.
+currently return explicit emulator-only 501 responses. See evidence for these limits.
 
 ## Idempotent submission and synthetic scope
 
@@ -149,7 +149,7 @@ The fingerprint covers the entire parsed JSON body, including unknown fields,
 and the trimmed `X-Hermes-Session-Key`. Object key order does not matter; array
 order and session/body changes do. Raw integer/float forms follow Python:
 `1` and `1.0` conflict, `1.0` and `1e0` replay, and negative floating zero differs
-from positive zero. Lone Unicode surrogates return an explicit mock-only 501;
+from positive zero. Lone Unicode surrogates return an explicit emulator-only 501;
 Python cannot UTF-8 encode that fingerprint either.
 
 `POST /__admin/hermes/scope` with `{ "profile": "synthetic-profile",
@@ -197,7 +197,7 @@ restores the owner flags and run/reservation state together.
 
 The controlled clock governs two independent retention layers. Cached `completed`,
 `failed` and `cancelled` results expire when their update age is strictly greater
-than one hour; equality survives and `interrupted` is excluded. The mock evaluates
+than one hour; equality survives and `interrupted` is excluded. The emulator evaluates
 elapsed 60-second sweep ticks on admission or observation, without a background
 process. `POST /__admin/hermes/sweep` with `{}` explicitly executes one sweep at the
 current clock, returning `cacheRemoved` and `simulated: true`.

package/SUPPORT.md

@@ -3,14 +3,14 @@
 Generated from `openapi.yaml`; do not edit by hand.
 
 - operations in spec: **6**
-- supported by the mock: **3**
+- supported by the emulator: **3**
 - parity enabled: **3**
 
-| operationId | route | mock | parity | notes |
+| operationId | route | emulator | parity | notes |
 | --- | --- | --- | --- | --- |
 | `RunCreate` | `POST /v1/runs` | ✅ supported | ⚠️ unsafe (opt-in) |  |
 | `RunGet` | `GET /v1/runs/{run_id}` | ✅ supported | ✅ |  |
 | `RunStop` | `POST /v1/runs/{run_id}/stop` | ✅ supported | ⚠️ unsafe (opt-in) |  |
-| `RunEvents` | `GET /v1/runs/{run_id}/events` | ❌ unsupported | — | SSE is outside the public peer-run mock subset. |
-| `RunApproval` | `POST /v1/runs/{run_id}/approval` | ❌ unsupported | — | Host tool approval execution is outside this mock. |
-| `RunSteer` | `POST /v1/runs/{run_id}/steer` | ❌ unsupported | — | Agent execution and steering are outside this mock. |
+| `RunEvents` | `GET /v1/runs/{run_id}/events` | ❌ unsupported | — | SSE is outside the public peer-run emulator subset. |
+| `RunApproval` | `POST /v1/runs/{run_id}/approval` | ❌ unsupported | — | Host tool approval execution is outside this emulator. |
+| `RunSteer` | `POST /v1/runs/{run_id}/steer` | ❌ unsupported | — | Agent execution and steering are outside this emulator. |