Compare commits
1 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
1cc34a8c31 |
@@ -1,69 +0,0 @@
|
||||
"""First-class context snapshot artifacts for live runtime memory evaluation."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from hermes_constants import get_hermes_home
|
||||
|
||||
|
||||
_SAFE_SEGMENT_RE = re.compile(r"[^A-Za-z0-9_.-]+")
|
||||
|
||||
|
||||
class ContextSnapshotRecorder:
|
||||
"""Write per-call prompt-composition artifacts for a Hermes session."""
|
||||
|
||||
def __init__(self, session_id: str, *, enabled: bool = False, base_dir: str | Path | None = None):
|
||||
self.session_id = session_id or "session"
|
||||
self.enabled = bool(enabled)
|
||||
self.base_dir = Path(base_dir) if base_dir else get_hermes_home() / "reports" / "context_snapshots"
|
||||
|
||||
@property
|
||||
def session_dir(self) -> Path:
|
||||
safe_session = _SAFE_SEGMENT_RE.sub("_", self.session_id).strip("._") or "session"
|
||||
return self.base_dir / safe_session
|
||||
|
||||
def record_call(
|
||||
self,
|
||||
api_call_count: int,
|
||||
*,
|
||||
system_prompt: str,
|
||||
memory_provider_system_prompt: str = "",
|
||||
memory_prefetch_raw: str = "",
|
||||
memory_context_block: str = "",
|
||||
api_user_message: str = "",
|
||||
api_messages: list[dict[str, Any]] | None = None,
|
||||
metadata: dict[str, Any] | None = None,
|
||||
) -> Path | None:
|
||||
if not self.enabled:
|
||||
return None
|
||||
|
||||
call_dir = self.session_dir / f"call_{api_call_count:03d}"
|
||||
call_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
self._write_text(call_dir / "system_prompt.txt", system_prompt or "")
|
||||
self._write_text(call_dir / "memory_provider_system_prompt.txt", memory_provider_system_prompt or "")
|
||||
self._write_text(call_dir / "memory_prefetch_raw.txt", memory_prefetch_raw or "")
|
||||
self._write_text(call_dir / "memory_context_block.txt", memory_context_block or "")
|
||||
self._write_text(call_dir / "api_user_message.txt", api_user_message or "")
|
||||
self._write_json(call_dir / "api_messages.json", api_messages or [])
|
||||
self._write_json(
|
||||
call_dir / "metadata.json",
|
||||
{
|
||||
"session_id": self.session_id,
|
||||
"api_call_count": api_call_count,
|
||||
**(metadata or {}),
|
||||
},
|
||||
)
|
||||
return call_dir
|
||||
|
||||
@staticmethod
|
||||
def _write_text(path: Path, content: str) -> None:
|
||||
path.write_text(content, encoding="utf-8")
|
||||
|
||||
@staticmethod
|
||||
def _write_json(path: Path, payload: Any) -> None:
|
||||
path.write_text(json.dumps(payload, indent=2, ensure_ascii=False), encoding="utf-8")
|
||||
@@ -1,132 +0,0 @@
|
||||
# Hindsight local eval homes for live Hermes runtime testing
|
||||
|
||||
Issue: #1010
|
||||
Parent: #985
|
||||
|
||||
This document defines a reproducible, profile-scoped evaluation layout for baseline / MemPalace / Hindsight comparisons without requiring Hindsight Cloud.
|
||||
|
||||
## Eval home layout
|
||||
|
||||
Use three separate `HERMES_HOME` directories so each run has isolated config, memory, sessions, and artifacts.
|
||||
|
||||
```text
|
||||
~/.hermes/profiles/atlas-baseline/
|
||||
config.yaml
|
||||
.env
|
||||
MEMORY.md
|
||||
USER.md
|
||||
reports/context_snapshots/
|
||||
|
||||
~/.hermes/profiles/atlas-mempalace/
|
||||
config.yaml
|
||||
.env
|
||||
MEMORY.md
|
||||
USER.md
|
||||
reports/context_snapshots/
|
||||
plugins/ # if a local MemPalace plugin is installed for this eval lane
|
||||
|
||||
~/.hermes/profiles/atlas-hindsight/
|
||||
config.yaml
|
||||
.env
|
||||
MEMORY.md
|
||||
USER.md
|
||||
hindsight/config.json
|
||||
reports/context_snapshots/
|
||||
```
|
||||
|
||||
## Hindsight local config
|
||||
|
||||
The Hindsight provider already loads config from `$HERMES_HOME/hindsight/config.json` first. For the local eval lane, prefer `local_embedded` so Hermes can bring up a local Hindsight daemon without cloud signup.
|
||||
|
||||
Example `~/.hermes/profiles/atlas-hindsight/hindsight/config.json`:
|
||||
|
||||
```json
|
||||
{
|
||||
"mode": "local_embedded",
|
||||
"memory_mode": "context",
|
||||
"recall_prefetch_method": "recall",
|
||||
"llm_provider": "ollama",
|
||||
"llm_model": "gemma3:12b",
|
||||
"api_url": "http://localhost:8888"
|
||||
}
|
||||
```
|
||||
|
||||
Notes:
|
||||
- `local_embedded` avoids any Hindsight Cloud dependency.
|
||||
- If `profile` is omitted, Hermes now derives a stable local Hindsight profile name from the active profile identity / `HERMES_HOME` instead of collapsing all local runs into the shared legacy `hermes` profile.
|
||||
- `local_external` remains valid if you already run a local Hindsight server yourself.
|
||||
|
||||
## Runtime switching procedure
|
||||
|
||||
Switch by exporting `HERMES_HOME` before launching Hermes.
|
||||
|
||||
### 1. Baseline
|
||||
|
||||
```bash
|
||||
export HERMES_HOME="$HOME/.hermes/profiles/atlas-baseline"
|
||||
unset HERMES_CONTEXT_SNAPSHOTS
|
||||
hermes chat
|
||||
```
|
||||
|
||||
### 2. MemPalace lane
|
||||
|
||||
```bash
|
||||
export HERMES_HOME="$HOME/.hermes/profiles/atlas-mempalace"
|
||||
export HERMES_CONTEXT_SNAPSHOTS=1
|
||||
hermes chat
|
||||
```
|
||||
|
||||
### 3. Hindsight lane
|
||||
|
||||
```bash
|
||||
export HERMES_HOME="$HOME/.hermes/profiles/atlas-hindsight"
|
||||
export HERMES_CONTEXT_SNAPSHOTS=1
|
||||
hermes chat
|
||||
```
|
||||
|
||||
## Raw artifact capture
|
||||
|
||||
When `HERMES_CONTEXT_SNAPSHOTS=1` is enabled, Hermes writes first-class prompt-composition artifacts under the active home by default.
|
||||
|
||||
Artifact tree:
|
||||
|
||||
```text
|
||||
$HERMES_HOME/reports/context_snapshots/<session-id>/call_001/
|
||||
system_prompt.txt
|
||||
memory_provider_system_prompt.txt
|
||||
memory_prefetch_raw.txt
|
||||
memory_context_block.txt
|
||||
api_user_message.txt
|
||||
api_messages.json
|
||||
metadata.json
|
||||
```
|
||||
|
||||
Minimum files a benchmark should inspect:
|
||||
- `system_prompt.txt`
|
||||
- `memory_prefetch_raw.txt`
|
||||
- `memory_context_block.txt`
|
||||
- `api_user_message.txt`
|
||||
- `api_messages.json`
|
||||
|
||||
These prove:
|
||||
- what the system prompt was
|
||||
- what the provider prefetched
|
||||
- what entered `<memory-context>`
|
||||
- what the final API user message looked like
|
||||
- what full payload reached the model
|
||||
|
||||
## Follow-on benchmark workflow
|
||||
|
||||
A benchmark issue can now consume this path without redoing integration work:
|
||||
1. pick one eval home (`atlas-baseline`, `atlas-mempalace`, `atlas-hindsight`)
|
||||
2. export the corresponding `HERMES_HOME`
|
||||
3. run Hermes on the same prompt set
|
||||
4. compare the snapshot artifacts in `reports/context_snapshots/`
|
||||
5. score recall quality and answer quality separately
|
||||
|
||||
## Why this is sovereign
|
||||
|
||||
- no hosted Hindsight Cloud dependency is required
|
||||
- the Hindsight config is profile-scoped under `hindsight/config.json`
|
||||
- the runtime artifacts stay under the active `HERMES_HOME`
|
||||
- switching between baseline / MemPalace / Hindsight is just a `HERMES_HOME` swap
|
||||
190
optional-skills/dogfood/adversarial-ux-test/SKILL.md
Normal file
190
optional-skills/dogfood/adversarial-ux-test/SKILL.md
Normal file
@@ -0,0 +1,190 @@
|
||||
---
|
||||
name: adversarial-ux-test
|
||||
description: Roleplay the most difficult, tech-resistant user for your product. Browse the app as that persona, find every UX pain point, then filter complaints through a pragmatism layer to separate real problems from noise. Creates actionable tickets from genuine issues only.
|
||||
version: 1.0.0
|
||||
author: Omni @ Comelse
|
||||
license: MIT
|
||||
metadata:
|
||||
hermes:
|
||||
tags: [qa, ux, testing, adversarial, dogfood, personas, user-testing]
|
||||
related_skills: [dogfood]
|
||||
---
|
||||
|
||||
# Adversarial UX Test
|
||||
|
||||
Roleplay the worst-case user for your product — the person who hates technology, doesn't want your software, and will find every reason to complain. Then filter their feedback through a pragmatism layer to separate real UX problems from "I hate computers" noise.
|
||||
|
||||
Think of it as an automated "mom test" — but angry.
|
||||
|
||||
## Why This Works
|
||||
|
||||
Most QA finds bugs. This finds **friction**. A technically correct app can still be unusable for real humans. The adversarial persona catches:
|
||||
- Confusing terminology that makes sense to developers but not users
|
||||
- Too many steps to accomplish basic tasks
|
||||
- Missing onboarding or "aha moments"
|
||||
- Accessibility issues (font size, contrast, click targets)
|
||||
- Cold-start problems (empty states, no demo content)
|
||||
- Paywall/signup friction that kills conversion
|
||||
|
||||
The **pragmatism filter** (Phase 3) is what makes this useful instead of just entertaining. Without it, you'd add a "print this page" button to every screen because Grandpa can't figure out PDFs.
|
||||
|
||||
## How to Use
|
||||
|
||||
Tell the agent:
|
||||
```
|
||||
"Run an adversarial UX test on [URL]"
|
||||
"Be a grumpy [persona type] and test [app name]"
|
||||
"Do an asshole user test on my staging site"
|
||||
```
|
||||
|
||||
You can provide a persona or let the agent generate one based on your product's target audience.
|
||||
|
||||
## Step 1: Define the Persona
|
||||
|
||||
If no persona is provided, generate one by answering:
|
||||
|
||||
1. **Who is the HARDEST user for this product?** (age 50+, non-technical role, decades of experience doing it "the old way")
|
||||
2. **What is their tech comfort level?** (the lower the better — WhatsApp-only, paper notebooks, wife set up their email)
|
||||
3. **What is the ONE thing they need to accomplish?** (their core job, not your feature list)
|
||||
4. **What would make them give up?** (too many clicks, jargon, slow, confusing)
|
||||
5. **How do they talk when frustrated?** (blunt, sweary, dismissive, sighing)
|
||||
|
||||
### Good Persona Example
|
||||
> **"Big Mick" McAllister** — 58-year-old S&C coach. Uses WhatsApp and that's it. His "spreadsheet" is a paper notebook. "If I can't figure it out in 10 seconds I'm going back to my notebook." Needs to log session results for 25 players. Hates small text, jargon, and passwords.
|
||||
|
||||
### Bad Persona Example
|
||||
> "A user who doesn't like the app" — too vague, no constraints, no voice.
|
||||
|
||||
The persona must be **specific enough to stay in character** for 20 minutes of testing.
|
||||
|
||||
## Step 2: Become the Asshole (Browse as the Persona)
|
||||
|
||||
1. Read any available project docs for app context and URLs
|
||||
2. **Fully inhabit the persona** — their frustrations, limitations, goals
|
||||
3. Navigate to the app using browser tools
|
||||
4. **Attempt the persona's ACTUAL TASKS** (not a feature tour):
|
||||
- Can they do what they came to do?
|
||||
- How many clicks/screens to accomplish it?
|
||||
- What confuses them?
|
||||
- What makes them angry?
|
||||
- Where do they get lost?
|
||||
- What would make them give up and go back to their old way?
|
||||
|
||||
5. Test these friction categories:
|
||||
- **First impression** — would they even bother past the landing page?
|
||||
- **Core workflow** — the ONE thing they need to do most often
|
||||
- **Error recovery** — what happens when they do something wrong?
|
||||
- **Readability** — text size, contrast, information density
|
||||
- **Speed** — does it feel faster than their current method?
|
||||
- **Terminology** — any jargon they wouldn't understand?
|
||||
- **Navigation** — can they find their way back? do they know where they are?
|
||||
|
||||
6. Take screenshots of every pain point
|
||||
7. Check browser console for JS errors on every page
|
||||
|
||||
## Step 3: The Rant (Write Feedback in Character)
|
||||
|
||||
Write the feedback AS THE PERSONA — in their voice, with their frustrations. This is not a bug report. This is a real human venting.
|
||||
|
||||
```
|
||||
[PERSONA NAME]'s Review of [PRODUCT]
|
||||
|
||||
Overall: [Would they keep using it? Yes/No/Maybe with conditions]
|
||||
|
||||
THE GOOD (grudging admission):
|
||||
- [things even they have to admit work]
|
||||
|
||||
THE BAD (legitimate UX issues):
|
||||
- [real problems that would stop them from using the product]
|
||||
|
||||
THE UGLY (showstoppers):
|
||||
- [things that would make them uninstall/cancel immediately]
|
||||
|
||||
SPECIFIC COMPLAINTS:
|
||||
1. [Page/feature]: "[quote in persona voice]" — [what happened, expected]
|
||||
2. ...
|
||||
|
||||
VERDICT: "[one-line persona quote summarizing their experience]"
|
||||
```
|
||||
|
||||
## Step 4: The Pragmatism Filter (Critical — Do Not Skip)
|
||||
|
||||
Step OUT of the persona. Evaluate each complaint as a product person:
|
||||
|
||||
- **RED: REAL UX BUG** — Any user would have this problem, not just grumpy ones. Fix it.
|
||||
- **YELLOW: VALID BUT LOW PRIORITY** — Real issue but only for extreme users. Note it.
|
||||
- **WHITE: PERSONA NOISE** — "I hate computers" talking, not a product problem. Skip it.
|
||||
- **GREEN: FEATURE REQUEST** — Good idea hidden in the complaint. Consider it.
|
||||
|
||||
### Filter Criteria
|
||||
1. Would a 35-year-old competent-but-busy user have the same complaint? → RED
|
||||
2. Is this a genuine accessibility issue (font size, contrast, click targets)? → RED
|
||||
3. Is this "I want it to work like paper" resistance to digital? → WHITE
|
||||
4. Is this a real workflow inefficiency the persona stumbled on? → YELLOW or RED
|
||||
5. Would fixing this add complexity for the 80% who are fine? → WHITE
|
||||
6. Does the complaint reveal a missing onboarding moment? → GREEN
|
||||
|
||||
**This filter is MANDATORY.** Never ship raw persona complaints as tickets.
|
||||
|
||||
## Step 5: Create Tickets
|
||||
|
||||
For **RED** and **GREEN** items only:
|
||||
- Clear, actionable title
|
||||
- Include the persona's verbatim quote (entertaining + memorable)
|
||||
- The real UX issue underneath (objective)
|
||||
- A suggested fix (actionable)
|
||||
- Tag/label: "ux-review"
|
||||
|
||||
For **YELLOW** items: one catch-all ticket with all notes.
|
||||
|
||||
**WHITE** items appear in the report only. No tickets.
|
||||
|
||||
**Max 10 tickets per session** — focus on the worst issues.
|
||||
|
||||
## Step 6: Report
|
||||
|
||||
Deliver:
|
||||
1. The persona rant (Step 3) — entertaining and visceral
|
||||
2. The filtered assessment (Step 4) — pragmatic and actionable
|
||||
3. Tickets created (Step 5) — with links
|
||||
4. Screenshots of key issues
|
||||
|
||||
## Tips
|
||||
|
||||
- **One persona per session.** Don't mix perspectives.
|
||||
- **Stay in character during Steps 2-3.** Break character only at Step 4.
|
||||
- **Test the CORE WORKFLOW first.** Don't get distracted by settings pages.
|
||||
- **Empty states are gold.** New user experience reveals the most friction.
|
||||
- **The best findings are RED items the persona found accidentally** while trying to do something else.
|
||||
- **If the persona has zero complaints, your persona is too tech-savvy.** Make them older, less patient, more set in their ways.
|
||||
- **Run this before demos, launches, or after shipping a batch of features.**
|
||||
- **Register as a NEW user when possible.** Don't use pre-seeded admin accounts — the cold start experience is where most friction lives.
|
||||
- **Zero WHITE items is a signal, not a failure.** If the pragmatism filter finds no noise, your product has real UX problems, not just a grumpy persona.
|
||||
- **Check known issues in project docs AFTER the test.** If the persona found a bug that's already in the known issues list, that's actually the most damning finding — it means the team knew about it but never felt the user's pain.
|
||||
- **Subscription/paywall testing is critical.** Test with expired accounts, not just active ones. The "what happens when you can't pay" experience reveals whether the product respects users or holds their data hostage.
|
||||
- **Count the clicks to accomplish the persona's ONE task.** If it's more than 5, that's almost always a RED finding regardless of persona tech level.
|
||||
|
||||
## Example Personas by Industry
|
||||
|
||||
These are starting points — customize for your specific product:
|
||||
|
||||
| Product Type | Persona | Age | Key Trait |
|
||||
|-------------|---------|-----|-----------|
|
||||
| CRM | Retirement home director | 68 | Filing cabinet is the current CRM |
|
||||
| Photography SaaS | Rural wedding photographer | 62 | Books clients by phone, invoices on paper |
|
||||
| AI/ML Tool | Department store buyer | 55 | Burned by 3 failed tech startups |
|
||||
| Fitness App | Old-school gym coach | 58 | Paper notebook, thick fingers, bad eyes |
|
||||
| Accounting | Family bakery owner | 64 | Shoebox of receipts, hates subscriptions |
|
||||
| E-commerce | Market stall vendor | 60 | Cash only, smartphone is for calls |
|
||||
| Healthcare | Senior GP | 63 | Dictates notes, nurse handles the computer |
|
||||
| Education | Veteran teacher | 57 | Chalk and talk, worksheets in ring binders |
|
||||
|
||||
## Rules
|
||||
|
||||
- Stay in character during Steps 2-3
|
||||
- Be genuinely mean but fair — find real problems, not manufactured ones
|
||||
- The pragmatism filter (Step 4) is **MANDATORY**
|
||||
- Screenshots required for every complaint
|
||||
- Max 10 tickets per session
|
||||
- Test on staging/deployed app, not local dev
|
||||
- One persona, one session, one report
|
||||
@@ -178,25 +178,6 @@ def _load_config() -> dict:
|
||||
}
|
||||
|
||||
|
||||
def _derive_local_profile_name(agent_identity: str = "", hermes_home: str = "") -> str:
|
||||
"""Return a stable profile name for local embedded Hindsight storage.
|
||||
|
||||
Prefer the active Hermes profile identity when available, otherwise fall back
|
||||
to the basename of the active HERMES_HOME path. This prevents all local
|
||||
Hindsight eval homes from sharing the legacy default profile name "hermes".
|
||||
"""
|
||||
from pathlib import Path
|
||||
import re
|
||||
|
||||
raw = (agent_identity or "").strip()
|
||||
if not raw and hermes_home:
|
||||
raw = Path(hermes_home).name.strip()
|
||||
if not raw:
|
||||
raw = "hermes"
|
||||
safe = re.sub(r"[^A-Za-z0-9_.-]+", "-", raw).strip(".-_")
|
||||
return safe or "hermes"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# MemoryProvider implementation
|
||||
# ---------------------------------------------------------------------------
|
||||
@@ -487,8 +468,6 @@ class HindsightMemoryProvider(MemoryProvider):
|
||||
|
||||
def initialize(self, session_id: str, **kwargs) -> None:
|
||||
self._session_id = session_id
|
||||
hermes_home = str(kwargs.get("hermes_home") or "")
|
||||
agent_identity = str(kwargs.get("agent_identity") or "")
|
||||
|
||||
# Check client version and auto-upgrade if needed
|
||||
try:
|
||||
@@ -521,11 +500,6 @@ class HindsightMemoryProvider(MemoryProvider):
|
||||
# "local" is a legacy alias for "local_embedded"
|
||||
if self._mode == "local":
|
||||
self._mode = "local_embedded"
|
||||
if self._mode == "local_embedded" and not self._config.get("profile"):
|
||||
self._config["profile"] = _derive_local_profile_name(
|
||||
agent_identity=agent_identity,
|
||||
hermes_home=hermes_home,
|
||||
)
|
||||
self._api_key = self._config.get("apiKey") or self._config.get("api_key") or os.environ.get("HINDSIGHT_API_KEY", "")
|
||||
default_url = _DEFAULT_LOCAL_URL if self._mode in ("local_embedded", "local_external") else _DEFAULT_API_URL
|
||||
self._api_url = self._config.get("api_url") or os.environ.get("HINDSIGHT_API_URL", default_url)
|
||||
|
||||
78
run_agent.py
78
run_agent.py
@@ -604,8 +604,6 @@ class AIAgent:
|
||||
checkpoint_max_snapshots: int = 50,
|
||||
pass_session_id: bool = False,
|
||||
persist_session: bool = True,
|
||||
context_snapshots_enabled: bool | None = None,
|
||||
context_snapshots_dir: str | None = None,
|
||||
):
|
||||
"""
|
||||
Initialize the AI Agent.
|
||||
@@ -1131,43 +1129,6 @@ class AIAgent:
|
||||
except Exception:
|
||||
_agent_cfg = {}
|
||||
|
||||
def _is_enabled(value):
|
||||
if isinstance(value, bool):
|
||||
return value
|
||||
return str(value).strip().lower() in {"1", "true", "yes", "on"}
|
||||
|
||||
_debug_cfg = _agent_cfg.get("debug", {}) if isinstance(_agent_cfg, dict) else {}
|
||||
if not isinstance(_debug_cfg, dict):
|
||||
_debug_cfg = {}
|
||||
_snapshot_cfg = _debug_cfg.get("context_snapshots", {})
|
||||
if not isinstance(_snapshot_cfg, dict):
|
||||
_snapshot_cfg = {}
|
||||
_snapshots_env = os.getenv("HERMES_CONTEXT_SNAPSHOTS")
|
||||
_snapshots_dir_env = os.getenv("HERMES_CONTEXT_SNAPSHOTS_DIR")
|
||||
if context_snapshots_enabled is None:
|
||||
if _snapshots_env is not None:
|
||||
self._context_snapshots_enabled = _is_enabled(_snapshots_env)
|
||||
else:
|
||||
self._context_snapshots_enabled = _is_enabled(_snapshot_cfg.get("enabled", False))
|
||||
else:
|
||||
self._context_snapshots_enabled = bool(context_snapshots_enabled)
|
||||
self._context_snapshots_dir = (
|
||||
context_snapshots_dir
|
||||
or _snapshots_dir_env
|
||||
or _snapshot_cfg.get("dir")
|
||||
or None
|
||||
)
|
||||
try:
|
||||
from agent.context_snapshots import ContextSnapshotRecorder
|
||||
self._context_snapshot_recorder = ContextSnapshotRecorder(
|
||||
session_id=self.session_id,
|
||||
enabled=self._context_snapshots_enabled,
|
||||
base_dir=self._context_snapshots_dir,
|
||||
)
|
||||
except Exception as _snapshot_err:
|
||||
logger.debug("Context snapshot recorder init failed: %s", _snapshot_err)
|
||||
self._context_snapshot_recorder = None
|
||||
|
||||
# Persistent memory (MEMORY.md + USER.md) -- loaded from disk
|
||||
self._memory_store = None
|
||||
self._memory_enabled = False
|
||||
@@ -8183,17 +8144,12 @@ class AIAgent:
|
||||
# Use original_user_message (clean input) — user_message may contain
|
||||
# injected skill content that bloats / breaks provider queries.
|
||||
_ext_prefetch_cache = ""
|
||||
_memory_provider_prompt_cache = ""
|
||||
if self._memory_manager:
|
||||
try:
|
||||
_query = original_user_message if isinstance(original_user_message, str) else ""
|
||||
_ext_prefetch_cache = self._memory_manager.prefetch_all(_query) or ""
|
||||
except Exception:
|
||||
pass
|
||||
try:
|
||||
_memory_provider_prompt_cache = self._memory_manager.build_system_prompt() or ""
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
while (api_call_count < self.max_iterations and self.iteration_budget.remaining > 0) or self._budget_grace_call:
|
||||
# Reset per-turn checkpoint dedup so each iteration can take one snapshot
|
||||
@@ -8261,8 +8217,6 @@ class AIAgent:
|
||||
# However, providers like Moonshot AI require a separate 'reasoning_content' field
|
||||
# on assistant messages with tool_calls. We handle both cases here.
|
||||
api_messages = []
|
||||
_current_api_user_message = ""
|
||||
_current_memory_context_block = ""
|
||||
for idx, msg in enumerate(messages):
|
||||
api_msg = msg.copy()
|
||||
|
||||
@@ -8277,15 +8231,12 @@ class AIAgent:
|
||||
_fenced = build_memory_context_block(_ext_prefetch_cache)
|
||||
if _fenced:
|
||||
_injections.append(_fenced)
|
||||
_current_memory_context_block = _fenced
|
||||
if _plugin_user_context:
|
||||
_injections.append(_plugin_user_context)
|
||||
if _injections:
|
||||
_base = api_msg.get("content", "")
|
||||
if isinstance(_base, str):
|
||||
api_msg["content"] = _base + "\n\n" + "\n\n".join(_injections)
|
||||
if isinstance(api_msg.get("content"), str):
|
||||
_current_api_user_message = api_msg["content"]
|
||||
|
||||
# For ALL assistant messages, pass reasoning back to the API
|
||||
# This ensures multi-turn reasoning context is preserved
|
||||
@@ -8320,13 +8271,7 @@ class AIAgent:
|
||||
from agent.privacy_filter import PrivacyFilter
|
||||
pf = PrivacyFilter()
|
||||
# Sanitize messages before they reach the provider
|
||||
_pf_result = pf.sanitize_messages(api_messages)
|
||||
if isinstance(_pf_result, tuple):
|
||||
api_messages, _pf_report = _pf_result
|
||||
if getattr(pf, "last_report", None) is None:
|
||||
pf.last_report = _pf_report
|
||||
else:
|
||||
api_messages = _pf_result
|
||||
api_messages = pf.sanitize_messages(api_messages)
|
||||
if pf.last_report and pf.last_report.had_redactions:
|
||||
logger.info(f"Privacy Filter: Redacted sensitive data from turn payload. Details: {pf.last_report.summary()}")
|
||||
except Exception as e:
|
||||
@@ -8397,27 +8342,6 @@ class AIAgent:
|
||||
new_tcs.append(tc)
|
||||
am["tool_calls"] = new_tcs
|
||||
|
||||
if self._context_snapshot_recorder:
|
||||
try:
|
||||
self._context_snapshot_recorder.record_call(
|
||||
api_call_count,
|
||||
system_prompt=effective_system,
|
||||
memory_provider_system_prompt=_memory_provider_prompt_cache,
|
||||
memory_prefetch_raw=_ext_prefetch_cache,
|
||||
memory_context_block=_current_memory_context_block,
|
||||
api_user_message=_current_api_user_message,
|
||||
api_messages=api_messages,
|
||||
metadata={
|
||||
"model": self.model,
|
||||
"provider": self.provider,
|
||||
"platform": self.platform or "",
|
||||
"api_mode": self.api_mode,
|
||||
"memory_providers": [p.name for p in getattr(self._memory_manager, "providers", [])],
|
||||
},
|
||||
)
|
||||
except Exception as _snapshot_err:
|
||||
logger.debug("Context snapshot capture failed: %s", _snapshot_err)
|
||||
|
||||
# Calculate approximate request size for logging
|
||||
total_chars = sum(len(str(msg)) for msg in api_messages)
|
||||
approx_tokens = estimate_messages_tokens_rough(api_messages)
|
||||
|
||||
@@ -1,43 +0,0 @@
|
||||
from pathlib import Path
|
||||
|
||||
from agent.context_snapshots import ContextSnapshotRecorder
|
||||
|
||||
|
||||
def test_disabled_recorder_writes_nothing(tmp_path):
|
||||
recorder = ContextSnapshotRecorder(session_id="session-1", enabled=False, base_dir=tmp_path)
|
||||
|
||||
out = recorder.record_call(
|
||||
1,
|
||||
system_prompt="system",
|
||||
api_messages=[{"role": "user", "content": "hello"}],
|
||||
)
|
||||
|
||||
assert out is None
|
||||
assert not (tmp_path / "session-1").exists()
|
||||
|
||||
|
||||
def test_enabled_recorder_writes_expected_artifacts(tmp_path):
|
||||
recorder = ContextSnapshotRecorder(session_id="session-1", enabled=True, base_dir=tmp_path)
|
||||
|
||||
out = recorder.record_call(
|
||||
1,
|
||||
system_prompt="system prompt",
|
||||
memory_provider_system_prompt="# Hindsight Memory\nActive.",
|
||||
memory_prefetch_raw="- remembered fact",
|
||||
memory_context_block="<memory-context>\nremembered\n</memory-context>",
|
||||
api_user_message="What do I prefer?\n\n<memory-context>\nremembered\n</memory-context>",
|
||||
api_messages=[
|
||||
{"role": "system", "content": "system prompt"},
|
||||
{"role": "user", "content": "What do I prefer?"},
|
||||
],
|
||||
metadata={"provider": "openai", "memory_providers": ["builtin", "hindsight"]},
|
||||
)
|
||||
|
||||
assert out == tmp_path / "session-1" / "call_001"
|
||||
assert (out / "system_prompt.txt").read_text(encoding="utf-8") == "system prompt"
|
||||
assert (out / "memory_provider_system_prompt.txt").read_text(encoding="utf-8").startswith("# Hindsight Memory")
|
||||
assert (out / "memory_prefetch_raw.txt").read_text(encoding="utf-8") == "- remembered fact"
|
||||
assert "<memory-context>" in (out / "memory_context_block.txt").read_text(encoding="utf-8")
|
||||
assert "What do I prefer?" in (out / "api_user_message.txt").read_text(encoding="utf-8")
|
||||
assert (out / "api_messages.json").read_text(encoding="utf-8").startswith("[")
|
||||
assert '"hindsight"' in (out / "metadata.json").read_text(encoding="utf-8")
|
||||
@@ -596,26 +596,3 @@ class TestAvailability:
|
||||
monkeypatch.setenv("HINDSIGHT_MODE", "local")
|
||||
p = HindsightMemoryProvider()
|
||||
assert p.is_available()
|
||||
|
||||
def test_local_embedded_profile_defaults_to_agent_identity(self, tmp_path, monkeypatch):
|
||||
config_path = tmp_path / "hindsight" / "config.json"
|
||||
config_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
config_path.write_text(json.dumps({
|
||||
"mode": "local_embedded",
|
||||
"llm_provider": "ollama",
|
||||
"llm_model": "gemma3:12b",
|
||||
}))
|
||||
monkeypatch.setattr(
|
||||
"plugins.memory.hindsight.get_hermes_home",
|
||||
lambda: tmp_path,
|
||||
)
|
||||
|
||||
p = HindsightMemoryProvider()
|
||||
p.initialize(
|
||||
session_id="test-session",
|
||||
hermes_home=str(tmp_path / "profiles" / "atlas-hindsight"),
|
||||
platform="cli",
|
||||
agent_identity="atlas-hindsight",
|
||||
)
|
||||
|
||||
assert p._config["profile"] == "atlas-hindsight"
|
||||
|
||||
@@ -1,94 +0,0 @@
|
||||
from pathlib import Path
|
||||
from types import SimpleNamespace
|
||||
from unittest.mock import MagicMock, patch
|
||||
import importlib
|
||||
import sys
|
||||
import types
|
||||
|
||||
|
||||
|
||||
def _make_tool_defs(*names: str) -> list:
|
||||
return [
|
||||
{
|
||||
"type": "function",
|
||||
"function": {
|
||||
"name": n,
|
||||
"description": f"{n} tool",
|
||||
"parameters": {"type": "object", "properties": {}},
|
||||
},
|
||||
}
|
||||
for n in names
|
||||
]
|
||||
|
||||
|
||||
|
||||
def _mock_response(content="Done", finish_reason="stop"):
|
||||
msg = SimpleNamespace(content=content, tool_calls=None)
|
||||
choice = SimpleNamespace(message=msg, finish_reason=finish_reason)
|
||||
return SimpleNamespace(choices=[choice], usage=SimpleNamespace(prompt_tokens=1, completion_tokens=1, total_tokens=2))
|
||||
|
||||
|
||||
|
||||
def _load_ai_agent():
|
||||
sys.modules.setdefault("agent.auxiliary_client", types.SimpleNamespace(call_llm=lambda *a, **k: ""))
|
||||
run_agent = importlib.import_module("run_agent")
|
||||
return run_agent.AIAgent
|
||||
|
||||
|
||||
|
||||
def test_run_conversation_writes_context_snapshot_artifacts(tmp_path):
|
||||
AIAgent = _load_ai_agent()
|
||||
|
||||
class _FakePrivacyFilter:
|
||||
def __init__(self):
|
||||
self.last_report = None
|
||||
|
||||
def sanitize_messages(self, messages):
|
||||
return list(messages)
|
||||
|
||||
with (
|
||||
patch("run_agent.get_tool_definitions", return_value=_make_tool_defs("web_search")),
|
||||
patch("run_agent.check_toolset_requirements", return_value={}),
|
||||
patch("run_agent.OpenAI"),
|
||||
patch("hermes_cli.plugins.invoke_hook", return_value=[]),
|
||||
patch.dict(sys.modules, {"agent.privacy_filter": types.SimpleNamespace(PrivacyFilter=_FakePrivacyFilter)}),
|
||||
):
|
||||
agent = AIAgent(
|
||||
api_key="test-key-1234567890",
|
||||
base_url="https://example.com/v1",
|
||||
quiet_mode=True,
|
||||
skip_context_files=True,
|
||||
skip_memory=True,
|
||||
context_snapshots_enabled=True,
|
||||
context_snapshots_dir=str(tmp_path),
|
||||
)
|
||||
|
||||
agent.client = MagicMock()
|
||||
agent.client.chat.completions.create.return_value = _mock_response(content="Done")
|
||||
agent._build_system_prompt = MagicMock(return_value="Core system prompt")
|
||||
agent._memory_manager = MagicMock()
|
||||
agent._memory_manager.prefetch_all.return_value = "- remembered preference"
|
||||
agent._memory_manager.build_system_prompt.return_value = "# Hindsight Memory\nActive."
|
||||
agent._memory_manager.providers = [
|
||||
SimpleNamespace(name="builtin"),
|
||||
SimpleNamespace(name="hindsight"),
|
||||
]
|
||||
|
||||
result = agent.run_conversation("What do I prefer?")
|
||||
|
||||
assert result["final_response"] == "Done"
|
||||
|
||||
call_dir = tmp_path / agent.session_id / "call_001"
|
||||
assert call_dir.exists()
|
||||
assert (call_dir / "system_prompt.txt").read_text(encoding="utf-8") == "Core system prompt"
|
||||
assert (call_dir / "memory_provider_system_prompt.txt").read_text(encoding="utf-8").startswith("# Hindsight Memory")
|
||||
assert (call_dir / "memory_prefetch_raw.txt").read_text(encoding="utf-8") == "- remembered preference"
|
||||
assert "<memory-context>" in (call_dir / "memory_context_block.txt").read_text(encoding="utf-8")
|
||||
api_user_message = (call_dir / "api_user_message.txt").read_text(encoding="utf-8")
|
||||
assert "What do I prefer?" in api_user_message
|
||||
assert "remembered preference" in api_user_message
|
||||
api_messages = (call_dir / "api_messages.json").read_text(encoding="utf-8")
|
||||
assert '"role": "system"' in api_messages
|
||||
assert '"role": "user"' in api_messages
|
||||
metadata = (call_dir / "metadata.json").read_text(encoding="utf-8")
|
||||
assert '"hindsight"' in metadata
|
||||
@@ -1,22 +0,0 @@
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
DOC = ROOT / "docs" / "hindsight-local-eval.md"
|
||||
|
||||
|
||||
def test_hindsight_local_eval_doc_exists_and_covers_switching():
|
||||
assert DOC.exists(), "missing Hindsight local eval doc"
|
||||
text = DOC.read_text(encoding="utf-8")
|
||||
for snippet in (
|
||||
"atlas-baseline",
|
||||
"atlas-mempalace",
|
||||
"atlas-hindsight",
|
||||
"HERMES_HOME",
|
||||
"HERMES_CONTEXT_SNAPSHOTS",
|
||||
"memory_prefetch_raw.txt",
|
||||
"api_user_message.txt",
|
||||
"local_embedded",
|
||||
"hindsight/config.json",
|
||||
):
|
||||
assert snippet in text
|
||||
25
tests/test_optional_adversarial_ux_skill_catalog.py
Normal file
25
tests/test_optional_adversarial_ux_skill_catalog.py
Normal file
@@ -0,0 +1,25 @@
|
||||
from pathlib import Path
|
||||
|
||||
from tools.skills_hub import OptionalSkillSource
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[1]
|
||||
|
||||
|
||||
def test_optional_skill_source_scans_adversarial_ux_test():
|
||||
source = OptionalSkillSource()
|
||||
metas = {meta.identifier: meta for meta in source._scan_all()}
|
||||
|
||||
assert "official/dogfood/adversarial-ux-test" in metas
|
||||
assert metas["official/dogfood/adversarial-ux-test"].name == "adversarial-ux-test"
|
||||
assert "tech-resistant user" in metas["official/dogfood/adversarial-ux-test"].description
|
||||
|
||||
|
||||
def test_optional_skill_catalog_docs_list_adversarial_ux_test():
|
||||
optional_catalog = (REPO_ROOT / "website" / "docs" / "reference" / "optional-skills-catalog.md").read_text(encoding="utf-8")
|
||||
bundled_catalog = (REPO_ROOT / "website" / "docs" / "reference" / "skills-catalog.md").read_text(encoding="utf-8")
|
||||
|
||||
assert "**adversarial-ux-test**" in optional_catalog
|
||||
assert "official/dogfood/adversarial-ux-test" in optional_catalog
|
||||
assert "`adversarial-ux-test`" in bundled_catalog
|
||||
assert "dogfood/adversarial-ux-test" in bundled_catalog
|
||||
@@ -16,6 +16,7 @@ For example:
|
||||
|
||||
```bash
|
||||
hermes skills install official/blockchain/solana
|
||||
hermes skills install official/dogfood/adversarial-ux-test
|
||||
hermes skills install official/mlops/flash-attention
|
||||
```
|
||||
|
||||
@@ -56,6 +57,12 @@ hermes skills uninstall <skill-name>
|
||||
| **blender-mcp** | Control Blender directly from Hermes via socket connection to the blender-mcp addon. Create 3D objects, materials, animations, and run arbitrary Blender Python (bpy) code. |
|
||||
| **meme-generation** | Generate real meme images by picking a template and overlaying text with Pillow. Produces actual `.png` meme files. |
|
||||
|
||||
## Dogfood
|
||||
|
||||
| Skill | Description |
|
||||
|-------|-------------|
|
||||
| **adversarial-ux-test** | Roleplay the most difficult, tech-resistant user for a product — browse in-persona, rant, then filter through a RED/YELLOW/WHITE/GREEN pragmatism layer so only real UX friction becomes tickets. |
|
||||
|
||||
## DevOps
|
||||
|
||||
| Skill | Description |
|
||||
|
||||
@@ -59,9 +59,12 @@ DevOps and infrastructure automation skills.
|
||||
|
||||
## dogfood
|
||||
|
||||
Internal dogfooding and QA skills used to test Hermes Agent itself.
|
||||
|
||||
| Skill | Description | Path |
|
||||
|-------|-------------|------|
|
||||
| `dogfood` | Systematic exploratory QA testing of web applications — find bugs, capture evidence, and generate structured reports. | `dogfood/dogfood` |
|
||||
| `adversarial-ux-test` | Roleplay the most difficult, tech-resistant user for a product — browse in-persona, rant, then filter through a RED/YELLOW/WHITE/GREEN pragmatism layer so only real UX friction becomes tickets. | `dogfood/adversarial-ux-test` |
|
||||
| `hermes-agent-setup` | Help users configure Hermes Agent — CLI usage, setup wizard, model/provider selection, tools, skills, voice/STT/TTS, gateway, and troubleshooting. | `dogfood/hermes-agent-setup` |
|
||||
|
||||
## email
|
||||
|
||||
Reference in New Issue
Block a user