Compare commits
1
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
 Bailey Dixon
|
0a5649d016 |
@@ -33,25 +33,26 @@ Chat goes directly to the API server via HTTP/SSE. The API key (Bearer token) is
|
||||
| `GET /health` | Health check | — |
|
||||
| `GET/POST/PATCH/DELETE /api/jobs/*` | Cron job management (api_server surface) | — |
|
||||
|
||||
**Non-standard endpoints (provided by fork OR by plugin bootstrap):**
|
||||
**Baseline upstream endpoints vs compatibility endpoints:**
|
||||
|
||||
These endpoints are not in stock upstream `gateway/platforms/api_server.py`. There are three ways a hermes-agent install can serve them:
|
||||
Upstream hermes-agent now has a native baseline for API Server session control and skill/toolset discovery:
|
||||
|
||||
1. **Codename-11 fork** (`feat/session-api` branch, deployed on the `axiom` branch) — adds them natively. Submitted upstream as PR [#8556](https://github.com/NousResearch/hermes-agent/pull/8556) *"feat(api-server): add session management API for frontend clients"* — scope is broader than the title: sessions CRUD + session chat/stream + memory + skills + config + available-models.
|
||||
2. **Bootstrap injection** (`hermes_relay_bootstrap/`) — monkey-patches aiohttp on startup via `.pth` file. Does NOT inject `/api/sessions/{id}/chat/stream` — use `/v1/runs` for chat.
|
||||
3. **Upstream-merged** (post PR #8556) — bootstrap auto-detects and no-ops.
|
||||
1. **Native upstream** — commit [`f7527b0`](https://github.com/NousResearch/hermes-agent/commit/f7527b0fdb54f01691547df03fc65a6d367f9fde), merged via PR [#33134](https://github.com/NousResearch/hermes-agent/pull/33134), salvaged the focused session-control work from closed PR [#29302](https://github.com/NousResearch/hermes-agent/pull/29302). It provides `/api/sessions/*`, session chat/stream, fork/messages, plus `/v1/skills` and `/v1/toolsets`.
|
||||
2. **Codename-11 `axiom` fork** — still carries compatibility/client-metadata routes that upstream does not provide yet: `/api/sessions/search`, `/api/memory`, `/api/skills` detail routes, `/api/config`, and `/api/available-models`.
|
||||
3. **Bootstrap injection** (`hermes_relay_bootstrap/`) — monkey-patches aiohttp on startup via `.pth` file and injects only missing compatibility routes for older or partial upstream builds. It should remain per-route/per-feature, not all-or-nothing.
|
||||
|
||||
| Endpoint | Purpose | Provided by |
|
||||
|----------|---------|-------------|
|
||||
| `GET /api/sessions` (CRUD) | Session list/create/rename/delete/fork | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET /api/sessions/{id}/messages` | Conversation history | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET /api/sessions/search` | Full-text message search | Fork OR bootstrap OR upstream-merged |
|
||||
| `POST /api/sessions/{id}/chat/stream` | Session-based SSE chat | Fork OR upstream-merged ONLY (NOT bootstrap) |
|
||||
| `GET /api/config`, `PATCH /api/config` | Personalities + model config | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET /api/skills`, `/{name}` | Skill discovery (list + detail) | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET /api/sessions` (CRUD) | Session list/create/rename/delete/fork | Native upstream OR fork OR bootstrap |
|
||||
| `GET /api/sessions/{id}/messages` | Conversation history | Native upstream OR fork OR bootstrap |
|
||||
| `GET /api/sessions/search` | Full-text message search | Fork OR bootstrap only |
|
||||
| `POST /api/sessions/{id}/chat/stream` | Session-based SSE chat | Native upstream OR fork only (NOT bootstrap) |
|
||||
| `GET /v1/skills` | Skill list metadata | Native upstream OR fork |
|
||||
| `GET /api/config`, `PATCH /api/config` | Personalities + model config | Fork OR bootstrap only |
|
||||
| `GET /api/skills`, `/{name}` | Legacy skill discovery/detail routes | Fork OR bootstrap only; Android prefers `/v1/skills` first |
|
||||
| `PUT /api/skills/toggle` | Enable/disable installed skill | `hermes_cli/web_server.py` dashboard surface; mirrored into bootstrap |
|
||||
| `GET/POST/PATCH/DELETE /api/memory` | Memory CRUD | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET /api/available-models` | Provider model list | Fork OR bootstrap OR upstream-merged |
|
||||
| `GET/POST/PATCH/DELETE /api/memory` | Memory CRUD | Fork OR bootstrap only |
|
||||
| `GET /api/available-models` | Provider-aware model list | Fork OR bootstrap only |
|
||||
|
||||
The Android client probes per-endpoint capability via `HermesApiClient.probeCapabilities()` (returns `ServerCapabilities`). When `streamingEndpoint = "auto"`, `ConnectionViewModel.resolveStreamingEndpoint()` picks `sessions` or `runs` based on the capability snapshot.
|
||||
|
||||
@@ -67,7 +68,7 @@ hermes-agent ships a second web server at `hermes_cli/web_server.py` that hosts
|
||||
## Key Instructions
|
||||
- **Always verify upstream before assuming an endpoint exists.** Check `gateway/platforms/api_server.py` in hermes-agent. If an endpoint isn't there, document whether bootstrap injects it or it requires the fork.
|
||||
- If we use a non-standard endpoint, ensure `probeCapabilities()` covers it and the auto-resolver degrades gracefully.
|
||||
- **Bootstrap maintenance:** Remove `hermes_relay_bootstrap/` in one PR once PR #8556 merges. It's no-op-compatible, so leaving it in place during rollout is harmless.
|
||||
- **Bootstrap maintenance:** Do not remove `hermes_relay_bootstrap/` just because upstream has native sessions. It can start shrinking only after each Relay-consuming compatibility route has a native replacement or the Android/Desktop clients have migrated away from it.
|
||||
|
||||
## Repository Layout
|
||||
|
||||
@@ -107,7 +108,7 @@ hermes-android/
|
||||
│ ├── tools/ # android_navigate.py, android_notifications.py
|
||||
│ └── dashboard/ # hermes-agent dashboard plugin — manifest, React UI, FastAPI proxy
|
||||
├── relay_server/ ← Thin compat shim → plugin.relay (legacy entrypoint)
|
||||
├── hermes_relay_bootstrap/ ← Runtime patch for vanilla upstream; removable after PR #8556
|
||||
├── hermes_relay_bootstrap/ ← Runtime patch for vanilla/partial upstream compatibility routes
|
||||
├── skills/devops/hermes-relay-pair/ ← /hermes-relay-pair slash command
|
||||
├── scripts/ ← dev.bat, bridge-smoke.sh, bump-version.sh
|
||||
└── docs/ ← spec, decisions, security, relay-server, mcp-tooling
|
||||
@@ -230,7 +231,7 @@ hermes-android/
|
||||
| `plugin/pair.py` | QR payload builder + CLI; `build_payload(sign=True)`; `--register-code` fallback |
|
||||
| `install.sh` | Canonical installer — 6 steps; idempotent; drops `hermes-relay-update` shim |
|
||||
| `uninstall.sh` | Canonical uninstaller; reverses install.sh; never touches `.env` or `state.db` |
|
||||
| `hermes_relay_bootstrap/` | Runtime patch for vanilla upstream; no-op on fork/upstream-merged; remove after PR #8556 |
|
||||
| `hermes_relay_bootstrap/` | Runtime patch for vanilla/partial upstream compatibility routes; shrink per route group after native parity or client migration |
|
||||
| **Plugin — Dashboard** | |
|
||||
| `plugin/dashboard/manifest.json` | Declares tab, entry bundle, and FastAPI module for hermes-agent discovery |
|
||||
| `plugin/dashboard/plugin_api.py` | FastAPI router proxying 5 routes to relay over loopback; `/pairing` body = API-server overrides (host/port/tls/api_key), relay URL auto-derived |
|
||||
|
||||
@@ -1,5 +1,15 @@
|
||||
# Hermes-Relay — Dev Log
|
||||
|
||||
## 2026-06-05 — Refresh upstream baseline docs after native session merge
|
||||
|
||||
**Context.** Upstream Hermes Agent merged native API Server session controls as commit `f7527b0` via PR #33134, and Android now prefers native `/v1/skills` with legacy fallback. Several Relay docs still treated PR #8556/#29302 as the future removal trigger for `hermes_relay_bootstrap/`.
|
||||
|
||||
**What changed.** Updated contributor/user docs to separate native upstream baseline routes (`/api/sessions/*`, `/v1/skills`, `/v1/toolsets`) from Relay compatibility routes that still require `axiom` or bootstrap (`/api/sessions/search`, `/api/memory`, `/api/config`, legacy skill detail routes, `/api/available-models`, voice aliases). The bootstrap retirement rule is now per route group, not wholesale deletion after one upstream session merge.
|
||||
|
||||
**Verification.** Ran stale-reference searches for `#8556`, `#29302`, and legacy skills routes after edits; remaining references are historical devlog/plans or explicitly marked compatibility/fallback.
|
||||
|
||||
---
|
||||
|
||||
## 2026-06-05 — Prefer upstream `/v1/skills` with legacy fallback
|
||||
|
||||
**Context.** Upstream Hermes Agent now has baseline skill/session API surface area, while Axiom's fork still preserves richer Relay-specific `/api/*` compatibility routes. The Android client should begin consuming upstream-compatible skill listings when present without breaking older fork/bootstrap installs.
|
||||
|
||||
+2
-2
@@ -101,7 +101,7 @@ Small follow-ons to v0.4 deliberately deferred to keep the v0.4.0 release surfac
|
||||
|
||||
**What the middleware can do (near-term, ships via install.sh).** New aiohttp middleware in `hermes_relay_bootstrap/_command_middleware.py`, installed at the same `_PatchedApplication.__setitem__` hook as the current route injection so it lands before `AppRunner.setup()` freezes the app. Filters by `request.path in ("/v1/runs", "/v1/chat/completions")` — zero-cost fast path for everything else. On chat paths: parses the body, lazy-imports `GATEWAY_KNOWN_COMMANDS` + `resolve_command()` + `gateway_help_lines()` from `hermes_cli.commands`, and splits on command type:
|
||||
- **Stateless commands** (`/help`, `/commands`, and any others the upstream Option B PR ends up supporting without router state) — actually dispatch, emit a synthetic SSE stream matching the runs handler's existing event shape so the Android client at `HermesApiClient.kt:655-715` renders it as a normal assistant turn.
|
||||
- **Stateful commands** (`/model`, `/new`, `/retry`, `/undo`, `/compress`, `/title`, `/resume`, `/branch`, `/rollback`, `/yolo`, `/reasoning`, `/personality`, etc. — most of the registry) — emit a synthetic SSE stream whose content is a short, helpful notice: *"The `/model` command requires a persistent session and isn't available on the stateless `/v1/runs` endpoint. Use `/api/sessions/{id}/chat/stream` (post-PR-#8556) or a channel with session state. For commands that work here, type `/help`."* This replaces the LLM hallucination with a deterministic, accurate message that points the user at the real fix.
|
||||
- **Stateful commands** (`/model`, `/new`, `/retry`, `/undo`, `/compress`, `/title`, `/resume`, `/branch`, `/rollback`, `/yolo`, `/reasoning`, `/personality`, etc. — most of the registry) — emit a synthetic SSE stream whose content is a short, helpful notice: *"The `/model` command requires a persistent session and isn't available on the stateless `/v1/runs` endpoint. Use `/api/sessions/{id}/chat/stream` or a channel with session state. For commands that work here, type `/help`."* This replaces the LLM hallucination with a deterministic, accurate message that points the user at the real fix.
|
||||
|
||||
**On no match** (unknown command, cli-only command, or plain text): falls through to `handler(request)` unchanged. Fork-detects the same way the existing injection does — if the upstream preprocessor PR lands first, the middleware no-ops.
|
||||
|
||||
@@ -109,7 +109,7 @@ Small follow-ons to v0.4 deliberately deferred to keep the v0.4.0 release surfac
|
||||
|
||||
**Files.** New `hermes_relay_bootstrap/_command_middleware.py` (~150 LOC), one-line append in `_patch.py` inside `_maybe_register_routes`, stdlib `unittest` coverage in `plugin/tests/test_bootstrap_command_middleware.py` mirroring the existing `test_bootstrap_patch.py` harness. Mirrors the upstream Option B PR exactly so the two can be reviewed side-by-side.
|
||||
|
||||
**Phase 2 — stateful dispatch on the session chat stream endpoint (post PR #8556).** Once PR #8556 merges and `/api/sessions/{id}/chat/stream` ships natively in upstream, a separate middleware (or a follow-up upstream PR) can add a preprocessor **scoped to that endpoint only**, leveraging the `session_id` in the URL as the persistence handle. At that point stateful commands become a dict write against session-scoped state — `session.model_override = new_model` — without needing to refactor `GatewayRouter` or plumb api_server into the router. Much smaller than a full router refactor, and it matches upstream's partition: `/v1/*` stays stateless, statefulness lives on `/api/sessions/*`. Blocked on #8556 landing.
|
||||
**Phase 2 — stateful dispatch on the session chat stream endpoint (unblocked by PR #33134 / commit `f7527b0`).** Since `/api/sessions/{id}/chat/stream` now ships natively in upstream, a separate middleware (or a follow-up upstream PR) can add a preprocessor **scoped to that endpoint only**, leveraging the `session_id` in the URL as the persistence handle. At that point stateful commands become a dict write against session-scoped state — `session.model_override = new_model` — without needing to refactor `GatewayRouter` or plumb api_server into the router. Much smaller than a full router refactor, and it matches upstream's partition: `/v1/*` stays stateless, statefulness lives on `/api/sessions/*`.
|
||||
|
||||
## Future — v0.5+
|
||||
|
||||
|
||||
@@ -78,10 +78,10 @@ Things to look into:
|
||||
- **Tool registration discoverability** — `android_*` tools register at gateway import time. There's no canonical "list installed plugin tools" API. Would adding one to upstream make sense, or is `gateway tool list` already enough?
|
||||
- **Versioning + compatibility ranges** — `pip install -e` doesn't enforce version pins between hermes-agent and our plugin. A breaking change in upstream's plugin loader could silently break us. Do we need a `hermes_compat: ">=0.8.0,<1.0.0"` field somewhere?
|
||||
- **`hermes-relay-self-setup` SKILL.md as a precedent** — we just shipped a self-installing skill that an LLM can fetch from a raw GitHub URL and execute. Does this pattern generalize? Could it become a recommended way for any third-party Hermes project to ship setup automation?
|
||||
- **Bootstrap injection** — `hermes_relay_bootstrap/` monkey-patches `aiohttp.web.Application` to inject endpoints into vanilla upstream. This is intentional but feels like a hack. Upstream PR #8556 (`feat/session-api`) will eventually let us delete it — verified 2026-04-15 that its scope covers the full bootstrap surface (sessions, memory, skills, config, available-models). Track that PR's status periodically.
|
||||
- **Gateway slash-command preprocessor — upstream Stage 1 PR.** Sibling follow-up to #8556. Intercepts known gateway commands on `/v1/runs` + `/v1/chat/completions`, dispatches the stateless ones (`/help`, `/commands`) via `gateway_help_lines()`, returns a deterministic "use a channel with session state" notice for the stateful majority. Currently being prepared in `C:/Users/Bailey/Desktop/Open-Projects/hermes-agent-pr-prep/` on branch `feat/api-server-gateway-commands`; awaiting subagent's code + draft PR body before pushing. See `docs/upstream-contributions.md` §5.
|
||||
- **Bootstrap injection shrink path** — `hermes_relay_bootstrap/` monkey-patches `aiohttp.web.Application` to inject endpoints into vanilla/partial upstream. Upstream commit `f7527b0` via PR #33134 now covers baseline sessions/chat/fork/message history, and `/v1/skills` covers list metadata. Do **not** delete the bootstrap wholesale yet: Relay still depends on compatibility routes that upstream lacks or does not match (`/api/sessions/search`, `/api/memory`, `/api/config`, legacy `/api/skills` detail routes, `/api/available-models`, and voice aliases). Shrink per route group only after native parity or client migration.
|
||||
- **Gateway slash-command preprocessor — upstream Stage 1 PR.** Follow-up to the native session-control baseline from PR #33134 / commit `f7527b0`. Intercepts known gateway commands on `/v1/runs` + `/v1/chat/completions`, dispatches the stateless ones (`/help`, `/commands`) via `gateway_help_lines()`, returns a deterministic "use a channel with session state" notice for the stateful majority. Currently being prepared in `C:/Users/Bailey/Desktop/Open-Projects/hermes-agent-pr-prep/` on branch `feat/api-server-gateway-commands`; awaiting subagent's code + draft PR body before pushing. See `docs/upstream-contributions.md` §5.
|
||||
- **Gateway slash-command preprocessor — bootstrap middleware (Stage 1 equivalent).** Sibling shim in `hermes_relay_bootstrap/_command_middleware.py` that mirrors the upstream Stage 1 PR as an aiohttp middleware injected at bootstrap time. Ships the hallucination fix to vanilla-upstream installs before the upstream PR lands. Planned for v0.4.1, after the current bridge feature branch wraps. See `ROADMAP.md` v0.4.1 entry.
|
||||
- **Stage 2 — stateful slash-command dispatch on `/api/sessions/{id}/chat/stream`.** Blocked on PR #8556 merging. Once session primitives ship upstream, add a preprocessor scoped to the session chat stream endpoint only, using `session_id` as the persistence handle. Separate upstream PR + matching bootstrap middleware. See `docs/upstream-contributions.md` §5 ("Stage 2").
|
||||
- **Stage 2 — stateful slash-command dispatch on `/api/sessions/{id}/chat/stream`.** Unblocked by upstream PR #33134 / commit `f7527b0`. Add a preprocessor scoped to the session chat stream endpoint only, using `session_id` as the persistence handle. Separate upstream PR + matching bootstrap middleware. See `docs/upstream-contributions.md` §5 ("Stage 2").
|
||||
|
||||
When the answer becomes clearer, this section becomes either an ADR in `docs/decisions.md` or a Plan under `Plans/`.
|
||||
|
||||
|
||||
@@ -259,8 +259,9 @@ GET /api/memory?target=memory // or target=user
|
||||
|
||||
### Skills
|
||||
```
|
||||
GET /api/skills # optional ?category= filter
|
||||
GET /api/skills/{name}
|
||||
GET /v1/skills # native upstream list metadata
|
||||
GET /api/skills # legacy fallback, optional ?category= filter
|
||||
GET /api/skills/{name} # legacy detail route when provided by fork/bootstrap
|
||||
```
|
||||
> `/api/skills/categories` was removed from upstream as dead code (commit 8d023e43) and is not re-injected by the bootstrap.
|
||||
|
||||
@@ -272,7 +273,8 @@ Probe endpoints to detect what's available:
|
||||
GET /health -> basic connectivity
|
||||
GET /api/sessions -> enhanced Hermes session API
|
||||
GET /v1/models -> model listing (OpenAI-compatible)
|
||||
GET /api/skills -> skills support
|
||||
GET /v1/skills -> native skills support
|
||||
GET /api/skills -> legacy skills fallback
|
||||
GET /api/memory -> memory support
|
||||
GET /api/config -> config API
|
||||
|
||||
|
||||
+34
-25
@@ -80,15 +80,19 @@ The app supports two streaming endpoints, selectable in Settings:
|
||||
| **Sessions** (`/api/sessions/{id}/chat/stream`) | Inline text annotations (`` `💻 terminal` ``) — client parses from markdown | Hermes-native SSE (assistant.delta, tool.progress, etc.) or OpenAI-format (delta.content) |
|
||||
| **Runs** (`/v1/runs` + `/v1/runs/{run_id}/events`) | **Structured events** (tool.started, tool.completed) — real-time tool cards | Hermes lifecycle events (message.delta, tool.started, tool.completed, run.completed) |
|
||||
|
||||
**Important upstream note:** The `/api/sessions` CRUD endpoints are moving
|
||||
toward upstream Hermes core through focused PR
|
||||
[#29302](https://github.com/NousResearch/hermes-agent/pull/29302), which covers
|
||||
**Important upstream note:** The `/api/sessions` CRUD/chat endpoints landed in
|
||||
upstream Hermes core via PR
|
||||
[#33134](https://github.com/NousResearch/hermes-agent/pull/33134) / commit
|
||||
[`f7527b0`](https://github.com/NousResearch/hermes-agent/commit/f7527b0fdb54f01691547df03fc65a6d367f9fde),
|
||||
which salvaged the focused PR
|
||||
[#29302](https://github.com/NousResearch/hermes-agent/pull/29302). It covers
|
||||
session list/create/read/update/delete, messages, fork, chat, and chat stream.
|
||||
Until that reaches a released core build, `hermes_relay_bootstrap/` still ships
|
||||
with the plugin and runs at Python interpreter startup via `.pth`. The bootstrap
|
||||
now composes with partial upstream support: native routes win per method/path,
|
||||
and the relay only injects missing compatibility gaps such as config, skills, or
|
||||
memory when core does not provide them.
|
||||
`hermes_relay_bootstrap/` still ships with the plugin and runs at Python
|
||||
interpreter startup via `.pth` for older cores and for compatibility gaps that
|
||||
upstream still does not provide. The bootstrap composes with partial upstream
|
||||
support: native routes win per method/path, and the relay only injects missing
|
||||
compatibility gaps such as search, config, legacy skills detail routes,
|
||||
available-models, or memory when core does not provide them.
|
||||
|
||||
The app's `probeCapabilities()` returns a per-endpoint snapshot, and `ConnectionViewModel.resolveStreamingEndpoint()` collapses `streamingEndpoint = "auto"` (the default for new installs) to a concrete `"sessions"` or `"runs"` choice based on what the server actually exposes.
|
||||
|
||||
@@ -499,11 +503,14 @@ Adopting from ARC's workflow patterns:
|
||||
### 16. Runtime API Server Patch via .pth Bootstrap (2026-04-12)
|
||||
|
||||
**Context:** The Android app depends on API-server routes for session history,
|
||||
profile/config metadata, skills, and memory-backed UI. Upstream core is now
|
||||
moving in focused pieces rather than one large frontend API patch: PR
|
||||
[#29302](https://github.com/NousResearch/hermes-agent/pull/29302) covers the
|
||||
canonical `/api/sessions/*` surface, while config/skills/memory still remain
|
||||
compatibility routes in this repo until core exposes stable equivalents. Without
|
||||
profile/config metadata, skills, and memory-backed UI. Upstream core has started
|
||||
absorbing the surface in focused pieces rather than one large frontend API patch:
|
||||
PR [#33134](https://github.com/NousResearch/hermes-agent/pull/33134) / commit
|
||||
[`f7527b0`](https://github.com/NousResearch/hermes-agent/commit/f7527b0fdb54f01691547df03fc65a6d367f9fde)
|
||||
now covers the canonical `/api/sessions/*` surface, and `/v1/skills` covers
|
||||
skill list metadata. Config, legacy skill detail routes, memory, search, and
|
||||
provider-aware model metadata still remain compatibility routes in this repo
|
||||
until core exposes stable equivalents or clients migrate away from them. Without
|
||||
the bootstrap, users on older vanilla upstream builds lose session browsing,
|
||||
metadata-backed settings, and history-on-restart behavior.
|
||||
|
||||
@@ -520,16 +527,17 @@ We considered four options:
|
||||
1. **Zero modifications to hermes-agent's filesystem.** `git pull` / `hermes update` see no local changes, so they always work cleanly. The patch lives entirely in `hermes_relay_bootstrap/` inside our own repo.
|
||||
2. **Single-file containment of all ported logic.** `_handlers.py` is 500 lines of straight-line aiohttp handler code with explicit `adapter` parameters (closures, not bound methods). Easy to audit, easy to delete.
|
||||
3. **Feature detection by method/path, not broad route family.** Native upstream
|
||||
routes win one method/path at a time. This matters because PR #29302 can land
|
||||
`/api/sessions/*` before core has stable config/skills/memory APIs; the
|
||||
bootstrap must not skip those remaining compatibility routes just because a
|
||||
sessions route exists.
|
||||
routes win one method/path at a time. This matters because PR #33134 landed
|
||||
`/api/sessions/*` before core had stable config/legacy skill detail/memory/search
|
||||
APIs; the bootstrap must not skip those remaining compatibility routes just
|
||||
because a sessions or `/v1/skills` route exists.
|
||||
4. **Trust model already established.** The user installed our plugin into their hermes-agent venv. They've already consented to having the plugin import hermes-agent internals (it does this for relay tools, voice endpoints, media registry). Monkey-patching `aiohttp.web.Application` is in the same trust bucket.
|
||||
5. **Surface-by-surface removal.** When PR #29302 or equivalent reaches a
|
||||
released hermes-agent version, the sessions compatibility routes should go
|
||||
quiet automatically. Config, skills, memory, and command preprocessing remain
|
||||
until their native replacements exist. Full bootstrap deletion happens only
|
||||
after every compatibility route group has a stable core equivalent.
|
||||
5. **Surface-by-surface removal.** When a supported hermes-agent baseline includes
|
||||
PR #33134 / `f7527b0`, the sessions compatibility routes should go quiet
|
||||
automatically. Config, legacy skills detail routes, memory, search,
|
||||
available-models, and command preprocessing remain until their native
|
||||
replacements exist or the clients migrate. Full bootstrap deletion happens
|
||||
only after every compatibility route group has a stable core equivalent.
|
||||
6. **`/v1/runs` is genuinely better for chat than `/api/sessions/{id}/chat/stream`.** It's standard upstream, supports `X-Hermes-Session-Id` for continuation, and emits live structured tool events. The fork's chat handler exists because upstream didn't HAVE this clean structured-event runs API at the time the fork was cut — but upstream does now.
|
||||
|
||||
**The Android client adapts via `streamingEndpoint = "auto"`.** New `ServerCapabilities` data class returned by `HermesApiClient.probeCapabilities()` captures per-endpoint presence (`sessionsApi`, `sessionsChatStream`, `runs`, `portable`, `healthy`). `ConnectionViewModel.resolveStreamingEndpoint()` collapses `"auto"` to `"sessions"` (when chat-stream handler is present, i.e. fork or upstream-merged) or `"runs"` (otherwise, i.e. bootstrap-injected vanilla upstream). The setting still supports manual `"sessions"` / `"runs"` overrides for debugging.
|
||||
@@ -548,10 +556,11 @@ We considered four options:
|
||||
- `install.sh` step 2 — copies the `.pth` into the venv site-packages
|
||||
|
||||
**Removal path** is now per surface:
|
||||
1. Sessions: after PR #29302 or equivalent ships in released core, keep the
|
||||
1. Sessions: with PR #33134 / `f7527b0` in the supported core baseline, keep the
|
||||
bootstrap installed but verify it skips native `/api/sessions/*` routes.
|
||||
2. Config/skills/memory: remove those compatibility handlers only after stable
|
||||
core APIs exist and Android probes prefer them.
|
||||
2. Config/legacy skills/memory/search/available-models: remove those
|
||||
compatibility handlers only after stable core APIs exist and Android probes
|
||||
prefer them, or after clients migrate away from the legacy route group.
|
||||
3. Slash middleware: remove after native API-server slash preprocessing exists.
|
||||
4. Full cleanup: delete `hermes_relay_bootstrap/`, delete
|
||||
`hermes_relay_bootstrap.pth`, remove the `.pth` install block, and update
|
||||
|
||||
@@ -293,8 +293,10 @@ which probes `hermes gateway run --help | grep tailscale`. When
|
||||
that returns true (PR #9295 has landed in your hermes-agent install),
|
||||
the helper still works but the canonical path
|
||||
(`hermes gateway run --tailscale`) is preferred and the helper will
|
||||
be removed in a future release. Same retirement pattern as
|
||||
`hermes_relay_bootstrap/` after PR #8556.
|
||||
be removed in a future release. Treat helper retirement like
|
||||
`hermes_relay_bootstrap/`: remove it only after the native path has parity
|
||||
for every Relay-consuming route/group, not merely after one upstream
|
||||
baseline lands.
|
||||
|
||||
### Forward-auth gateways (Authelia, Cloudflare Access) in front of the API server
|
||||
|
||||
|
||||
+3
-3
@@ -361,7 +361,7 @@ Bottom navigation bar with 4 tabs:
|
||||
3. Remaining top-bar actions (session drawer hamburger, ambient toggle, etc.).
|
||||
- **Session drawer** (swipe from left or hamburger icon) — session list with title, timestamp, message count. Create, switch, rename, delete.
|
||||
- **Chat view** — message bubbles with markdown rendering, streaming text, tool call cards (Off/Compact/Detailed display modes)
|
||||
- **Input bar** — text field with 4096 char limit, `/` palette button, send button, stop button during streaming. Inline autocomplete on `/` keystroke + full searchable command palette (bottom sheet). Commands sourced from: 29 gateway built-ins, dynamic personalities from `config.agent.personalities`, and server skills from `GET /api/skills`.
|
||||
- **Input bar** — text field with 4096 char limit, `/` palette button, send button, stop button during streaming. Inline autocomplete on `/` keystroke + full searchable command palette (bottom sheet). Commands sourced from: 29 gateway built-ins, dynamic personalities from `config.agent.personalities`, and server skills from `GET /v1/skills` with legacy `/api/skills` fallback.
|
||||
- **Empty state** — Logo + "Start a conversation" + suggestion chips that populate input
|
||||
- **Agent sheet — Profile section (v0.6.0, updated 2026-05-18)** — upstream Hermes profiles auto-discovered by the relay at `~/.hermes/profiles/*/`. Selecting one routes chat/session calls to that profile's advertised `api_server_url` when present, giving proper Hermes isolation for sessions, memory, tools, provider auth, and SOUL/default model. If no profile API route is advertised, the app falls back to overlaying `model` + `SOUL.md` (as `system_message`) on the active Connection's API server. Selection is persisted per Connection/profile context. Hidden when the server advertises no profiles. See `docs/decisions.md` §21.
|
||||
- **Agent sheet — Personality section** — personalities fetched from `GET /api/config` (`config.agent.personalities`). Shows server default (from `config.display.personality`) + all configured. Active personality name shown on assistant chat bubbles.
|
||||
@@ -499,7 +499,7 @@ Additional API endpoints used:
|
||||
5. DELETE /api/sessions/{session_id} → delete session
|
||||
6. GET /api/sessions/{session_id}/messages → fetch message history
|
||||
7. GET /api/config → personalities (for personality picker, `config.agent.personalities`)
|
||||
8. GET /api/skills → available skills (for command palette + autocomplete)
|
||||
8. GET /v1/skills → available skills (for command palette + autocomplete; fallback GET /api/skills)
|
||||
```
|
||||
|
||||
Key classes:
|
||||
@@ -888,7 +888,7 @@ Current versions as of v0.3.0. Source of truth is `gradle/libs.versions.toml`
|
||||
| **WebAPI chat** | HTTP to `localhost:8642/api/sessions/*/chat/stream` (SSE) |
|
||||
| **WebAPI sessions** | `GET/POST/PATCH/DELETE /api/sessions` for CRUD |
|
||||
| **Personalities** | `GET /api/config` → `config.agent.personalities` for picker + command palette |
|
||||
| **Server skills** | `GET /api/skills` — dynamic skill discovery for command palette + autocomplete |
|
||||
| **Server skills** | `GET /v1/skills` first, fallback `GET /api/skills` — dynamic skill discovery for command palette + autocomplete |
|
||||
| **Plugin system** | `register_tool()` via `ctx` for `android_*` tools |
|
||||
| **Gateway** | Chat channel goes through WebAPI, not directly to gateway |
|
||||
| **Memory/Skills** | Accessible through agent chat (no direct API needed for MVP) |
|
||||
|
||||
@@ -2,9 +2,11 @@
|
||||
|
||||
Improvements that would benefit hermes-relay (and other frontends) if added to [NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent).
|
||||
|
||||
## Current Upstream PR Alignment
|
||||
## Current Upstream Alignment
|
||||
|
||||
- PR #29302 (`feat: add API server session controls`) is the canonical upstream path for `/api/sessions/*`, message history, fork, chat, and chat stream. Hermes-Relay should prefer these native routes when present and keep the bootstrap as a per-route compatibility overlay only for older or partial core builds.
|
||||
- PR #33134 (`feat(api-server): session control API — sessions/chat/fork/SSE-stream (salvages #29302)`) landed upstream as commit [`f7527b0`](https://github.com/NousResearch/hermes-agent/commit/f7527b0fdb54f01691547df03fc65a6d367f9fde). It is now the canonical baseline for `/api/sessions/*`, message history, fork, chat, and chat stream.
|
||||
- `GET /v1/skills` and `GET /v1/toolsets` are native upstream capability-discovery routes. Hermes-Relay's Android client prefers `/v1/skills` first and falls back to legacy `/api/skills` for older/fork/bootstrap installs.
|
||||
- Keep `hermes_relay_bootstrap` as a per-route compatibility overlay only for older or partial core builds. Do not remove it wholesale while Relay still needs routes without native upstream parity: `/api/sessions/search`, `/api/memory`, `/api/config`, legacy skill detail routes under `/api/skills`, `/api/available-models`, and voice aliases.
|
||||
- PR #8199 (`feat(api): add native audio transcription and speech endpoints`) is the canonical upstream path for core STT/TTS execution through `/v1/audio/transcriptions` and `/v1/audio/speech`. Hermes-Relay should keep `/voice/*` as the paired-device facade but eventually call those native core endpoints internally before falling back to private helper imports.
|
||||
- PR #29364 (`feat: add API server audio endpoints`) should not become a competing `/api/audio/*` API if #8199 remains the accepted audio base. Rework it as a discovery/compatibility follow-up or close it after confirming the upstream maintainer preference.
|
||||
|
||||
@@ -38,7 +40,7 @@ Improvements that would benefit hermes-relay (and other frontends) if added to [
|
||||
|
||||
**Impact:** All frontends (hermes-relay, hermes-workspace, ClawPort) could dynamically show available commands without hardcoding. New commands added upstream would appear automatically.
|
||||
|
||||
**Workaround (current):** 29 gateway commands hardcoded in `ChatScreen.kt`, manually synced with `hermes_cli/commands.py`. Personality commands generated from `GET /api/config`. Skills from `GET /api/skills`.
|
||||
**Workaround (current):** 29 gateway commands hardcoded in `ChatScreen.kt`, manually synced with `hermes_cli/commands.py`. Personality commands generated from `GET /api/config`. Skills prefer native `GET /v1/skills` and fall back to legacy `GET /api/skills`.
|
||||
|
||||
## 2. Personality Switching via Dedicated API Parameter
|
||||
|
||||
@@ -103,16 +105,16 @@ except Exception as _exc:
|
||||
|
||||
**Proposed — a two-stage arc, each stage a small, independently reviewable PR:**
|
||||
|
||||
**Stage 1 — stateless preprocessor (sibling follow-up to PR #29302).** A lightweight preprocessor in `api_server.py`'s `/v1/runs` + `/v1/chat/completions` handlers that detects a leading `/` in the user text, matches the first token against `GATEWAY_KNOWN_COMMANDS`, and splits on command type:
|
||||
**Stage 1 — stateless preprocessor (follow-up to PR #33134 / commit `f7527b0`).** A lightweight preprocessor in `api_server.py`'s `/v1/runs` + `/v1/chat/completions` handlers that detects a leading `/` in the user text, matches the first token against `GATEWAY_KNOWN_COMMANDS`, and splits on command type:
|
||||
|
||||
- **Stateless commands** (`/help`, `/commands`, and any others that can execute without touching router-owned state) are dispatched via existing helpers (`gateway_help_lines()` at `hermes_cli/commands.py:340`) and returned as a synthetic SSE stream matching the handlers' existing event shape.
|
||||
- **Stateful commands** (`/model`, `/new`, `/retry`, `/undo`, `/compress`, `/title`, `/resume`, `/branch`, `/rollback`, `/yolo`, `/reasoning`, `/personality`, and most of the registry) return a deterministic, helpful SSE notice along the lines of *"The `/model` command requires a persistent session and isn't available on the stateless `/v1/runs` endpoint. Use `/api/sessions/{id}/chat/stream` (from PR #29302) or a channel with session state (Discord, CLI, Telegram)."*
|
||||
- **Stateful commands** (`/model`, `/new`, `/retry`, `/undo`, `/compress`, `/title`, `/resume`, `/branch`, `/rollback`, `/yolo`, `/reasoning`, `/personality`, and most of the registry) return a deterministic, helpful SSE notice along the lines of *"The `/model` command requires a persistent session and isn't available on the stateless `/v1/runs` endpoint. Use `/api/sessions/{id}/chat/stream` or a channel with session state (Discord, CLI, Telegram)."*
|
||||
- **Unknown** and **cli-only** commands fall through to the LLM path unchanged.
|
||||
- **Preprocessor exceptions** fall through to the LLM path unchanged — a preprocessor bug must never take down a normal chat request.
|
||||
|
||||
This respects upstream's intentional design (api_server stays stateless, no router coupling) while fixing the hallucination symptom and unlocking the commands that *can* run statelessly.
|
||||
|
||||
**Stage 2 — stateful dispatch on `/api/sessions/{id}/chat/stream` (after PR #29302 lands).** Once session management primitives ship, a separate PR adds a preprocessor **scoped to the session chat stream endpoint only**, using the URL's `session_id` as the persistence handle. Stateful commands become session-scoped dict writes (`session.model_override = new_model`) without refactoring `GatewayRouter` or plumbing api_server into the router. This matches upstream's partition cleanly: `/v1/*` remains stateless and OpenAI-compatible; statefulness lives on `/api/sessions/*`.
|
||||
**Stage 2 — stateful dispatch on `/api/sessions/{id}/chat/stream` (now unblocked).** Now that session management primitives have landed upstream, a separate PR can add a preprocessor **scoped to the session chat stream endpoint only**, using the URL's `session_id` as the persistence handle. Stateful commands become session-scoped dict writes (`session.model_override = new_model`) without refactoring `GatewayRouter` or plumbing api_server into the router. This matches upstream's partition cleanly: `/v1/*` remains stateless and OpenAI-compatible; statefulness lives on `/api/sessions/*`.
|
||||
|
||||
**Why not one big PR:** a full GatewayRouter refactor plus api_server plumbing was considered and rejected. It would touch 10+ files across subsystems normally owned separately, fight the documented "api_server is excluded from router notification" design decision, and review as a much larger change than the value added. The two-stage arc ships faster, reviews cleaner, and matches the upstream partition better.
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
# Upstream Hermes Integration Sync
|
||||
|
||||
Last reviewed: 2026-05-20
|
||||
Last reviewed: 2026-06-05
|
||||
|
||||
This document tracks how Hermes-Relay integrates with Hermes upstream surfaces, which
|
||||
parts use supported extension points, and which parts are compatibility layers that
|
||||
@@ -37,8 +37,8 @@ relay, dashboard, Android app, desktop app, bootstrap package, or user docs.
|
||||
| Agent tools | Tool Gateway tools registered through plugin context | `ctx.register_tool(...)` in `plugin/__init__.py`; schemas and handlers in `plugin/tools/*` | Aligned with custom transports | Tool registration should stay in `register(ctx)`; transport details stay behind handlers. |
|
||||
| Dashboard tab and plugin API | Dashboard plugin manifest plus plugin API routes under the Hermes dashboard plugin mount | `plugin/dashboard/manifest.json`, `plugin/dashboard/plugin_api.py` | Aligned wrapper | Dashboard routes may proxy relay state, but discovery and mounting should stay upstream-native. |
|
||||
| Chat and model API | OpenAI-compatible API server routes such as `/v1/chat/completions`, `/v1/models`, `/health`, and supported streaming routes | Android `HermesApiClient`, relay docs, Web API docs | Mixed | Prefer standard API routes first; use `/api/sessions` only when capability probes find it. |
|
||||
| Sessions API | Proposed upstream API-server session controls in NousResearch/hermes-agent PR #29302 (`/api/sessions`, messages, fork, chat, chat stream) | Android `HermesApiClient`; compatibility overlay in `hermes_relay_bootstrap/*` | Upstream-pending with fallback | Prefer native `/api/sessions/*` when present. Bootstrap must skip native routes per method/path and only inject missing compatibility routes. |
|
||||
| Config, skills, memory APIs | Not documented as stable upstream API-server routes in current public docs | `hermes_relay_bootstrap/*`, `docs/HERMES-WEBAPI-REFERENCE.md` | Compatibility layer | Keep separate from the sessions retirement path. Do not skip these just because native `/api/sessions` exists. |
|
||||
| Sessions API | Native upstream API-server session controls from PR #33134 / commit `f7527b0` (`/api/sessions`, messages, fork, chat, chat stream) | Android `HermesApiClient`; compatibility overlay in `hermes_relay_bootstrap/*` for older cores | Upstream baseline with fallback | Prefer native `/api/sessions/*` when present. Bootstrap must skip native routes per method/path and only inject missing compatibility routes. |
|
||||
| Config, skills, memory APIs | Native upstream provides `/v1/skills` and `/v1/toolsets`; config, memory, legacy skill detail routes, search, and available-models are not stable upstream API-server routes | `hermes_relay_bootstrap/*`, `docs/HERMES-WEBAPI-REFERENCE.md` | Mixed: native skill list + compatibility layer | Keep separate from the sessions retirement path. Do not skip these just because native `/api/sessions` or `/v1/skills` exists. |
|
||||
| Mobile, desktop, and terminal relay transport | No general upstream plugin WSS transport for persistent remote clients in current public docs | `plugin/relay/server.py`, `plugin/relay/channels/*` | Custom | Keep the relay protocol documented and avoid leaking relay-only assumptions into upstream API clients. |
|
||||
| Pairing QR and relay session minting | No upstream pairing or device-registration method for remote mobile clients in current public docs | `plugin/pair.py`, relay `/pairing/*`, Android QR parser | Custom | QR payloads should keep API credentials (`key`) separate from relay credentials (`relay.code`). |
|
||||
| Basic STT/TTS over HTTP | Proposed upstream API-server audio endpoints in PR #8199 (`/v1/audio/transcriptions`, `/v1/audio/speech`) | Relay `/voice/config`, `/voice/transcribe`, `/voice/synthesize`; Android `RelayVoiceClient`; `plugin/relay/upstream_voice.py` | Custom wrapper pending upstream replacement | Keep `/voice/*` as the relay auth/session compatibility facade. Once core audio endpoints land, prefer proxying to native `/v1/audio/*` for STT/TTS work before falling back to private helper imports. |
|
||||
@@ -50,7 +50,7 @@ relay, dashboard, Android app, desktop app, bootstrap package, or user docs.
|
||||
|
||||
| Deviation | Owner files | Why it exists | Guard or fallback | Retirement condition |
|
||||
| --- | --- | --- | --- | --- |
|
||||
| API bootstrap route and middleware injection | `hermes_relay_bootstrap/*` | Native installs need session/config/skills/memory endpoints and slash-command preprocessing before upstream exposes stable equivalents. | Method/path feature detection skips native upstream routes and injects only missing compatibility gaps; upstream-module checks skip middleware when native slash preprocessing exists. | Retire per surface: sessions after PR #29302 or equivalent ships in a released core; config/skills/memory after stable core APIs exist; slash middleware after native preprocessing exists. |
|
||||
| API bootstrap route and middleware injection | `hermes_relay_bootstrap/*` | Native/older installs need compatibility routes and slash-command preprocessing before upstream exposes stable equivalents. | Method/path feature detection skips native upstream routes and injects only missing compatibility gaps; upstream-module checks skip middleware when native slash preprocessing exists. | Retire per surface: sessions after the supported baseline includes PR #33134/`f7527b0` and tests prove no Relay session route gaps remain; config/legacy skills/memory/search/available-models after stable core APIs or client migrations exist; slash middleware after native preprocessing exists. |
|
||||
| Plugin CLI shim fallback | `plugin/__init__.py`, `plugin/cli.py`, install scripts | Some Hermes versions do not wire third-party plugin CLI commands into the top-level parser. | `ctx.register_cli_command` is attempted first; standalone shims fill the gap. | Remove shims once upstream plugin CLI discovery is stable for native installs. |
|
||||
| Relay HTTP and WSS server | `plugin/relay/server.py`, `plugin/relay/channels/*` | Mobile, desktop, terminal, media, push, and bridge features need persistent client channels and relay-owned session state. | Keep upstream API calls separate from relay session calls and document the protocol in `docs/relay-protocol.md`. | Replace pieces only when upstream provides equivalent remote-client transport or platform adapters. |
|
||||
| Pairing schema with `relay.code` | `plugin/pair.py`, Android pairing parser, relay `/pairing/*` | An API bearer key authenticates Hermes API calls but does not create relay sessions or describe WSS endpoints. | QR payloads carry direct API credentials and relay credentials as separate families. | Remove custom pairing when upstream offers native remote-device registration and relay discovery. |
|
||||
@@ -118,7 +118,7 @@ upgrading the supported Hermes baseline.
|
||||
- `GET /api/sessions?limit=1` only as an enhanced-management capability probe
|
||||
- Relay health and info endpoints from `docs/relay-server.md`
|
||||
- Dashboard plugin overview under the Hermes plugin API mount
|
||||
- `GET /v1/capabilities` and the native `/api/sessions/*` route set when testing a core build with PR #29302 or equivalent
|
||||
- `GET /v1/capabilities`, `GET /v1/skills`, and the native `/api/sessions/*` route set when testing a core build with PR #33134 / commit `f7527b0` or equivalent
|
||||
- `POST /v1/audio/transcriptions` and `POST /v1/audio/speech` when testing a core build with PR #8199 or equivalent
|
||||
- Voice config, transcription, synthesis, and realtime routes only with relay session auth or a valid Hermes API bearer
|
||||
6. Update this file when upstream adds a supported replacement for a custom layer.
|
||||
|
||||
@@ -12,7 +12,7 @@ The plugin is a thin observer — it never modifies state, never writes to your
|
||||
|
||||
**On your server:**
|
||||
|
||||
- hermes-agent with the Dashboard Plugin System (upstream commit `01214a7f` on `axiom`, or any later `main` once [PR #8556](https://github.com/NousResearch/hermes-agent/pull/8556) and its dashboard followups merge). `hermes dashboard start` must already work for you.
|
||||
- hermes-agent with the Dashboard Plugin System (upstream commit `01214a7f` on `axiom`, or any later `main` that includes the dashboard plugin follow-ups). `hermes dashboard start` must already work for you.
|
||||
- The canonical Hermes-Relay install — if you ran the one-liner on the [Quick Start](/guide/getting-started), you're done. The installer symlinks `~/.hermes/plugins/hermes-relay` → the plugin subtree and the dashboard scanner picks up `plugin/dashboard/manifest.json` automatically.
|
||||
- A gateway restart after install: `systemctl --user restart hermes-gateway`.
|
||||
|
||||
|
||||
@@ -55,7 +55,7 @@ All commands are fetched dynamically from the server where possible:
|
||||
- **Configuration**: `/model`, `/personality`, `/reasoning`, `/yolo`, `/verbose`, `/voice`
|
||||
- **Info**: `/help`, `/status`, `/usage`, `/insights`, `/commands`
|
||||
- **Personalities**: generated from server config (`config.agent.personalities`) — `/personality victor`, `/personality creative`, etc.
|
||||
- **Skills**: dynamically fetched from `GET /api/skills` — 90+ server skills grouped by category (creative, devops, research, etc.)
|
||||
- **Skills**: dynamically fetched from `GET /v1/skills` with fallback to legacy `GET /api/skills` — server skills grouped by category (creative, devops, research, etc.)
|
||||
|
||||
## Tool Execution
|
||||
|
||||
|
||||
@@ -18,9 +18,9 @@ Hermes-Relay voice endpoints can reuse this same Bearer token. Relay validates i
|
||||
|
||||
## How endpoints get served
|
||||
|
||||
Installing the plugin via `install.sh` is enough to make all of the endpoints below work — including the management ones (`/api/sessions/*`, `/api/memory`, `/api/skills`, `/api/config`, `/api/available-models`). The plugin wires the gateway up at install time so these are served on the same `:8642` host as the standard `/v1/*` endpoints, with the same `Authorization: Bearer …` auth.
|
||||
Current upstream hermes-agent includes the baseline session API (`/api/sessions/*`, chat, stream, fork, messages) and native capability discovery (`/v1/skills`, `/v1/toolsets`). Installing the Relay plugin via `install.sh` keeps older or partial core builds compatible by injecting only the missing management endpoints Relay still consumes, such as `/api/sessions/search`, `/api/memory`, `/api/config`, legacy `/api/skills` detail routes, `/api/available-models`, and voice aliases. These are served on the same `:8642` host as the standard `/v1/*` endpoints, with the same optional `Authorization: Bearer ***` auth.
|
||||
|
||||
**Chat streaming uses standard `/v1/runs`** by default — it emits structured `tool.started`/`tool.completed` SSE events for live tool progress cards in the Android app. The app's `Settings → Chat → Streaming endpoint = "Auto"` (default) probes per-endpoint capability and picks the best chat path automatically; you can manually force `Sessions` or `Runs` mode for debugging.
|
||||
**Chat streaming uses standard `/v1/runs`** by default — it emits structured `tool.started`/`tool.completed` SSE events for live tool progress cards in the Android app. The app's `Settings → Chat → Streaming endpoint = "Auto"` (default) probes per-endpoint capability and picks the best chat path automatically; you can manually force `Sessions` or `Runs` mode for debugging. Skill discovery prefers `GET /v1/skills` and falls back to legacy `GET /api/skills`.
|
||||
|
||||
## Endpoints
|
||||
|
||||
|
||||
Reference in New Issue
Block a user