Sami/hermes-agent
1
0
Fork 0
forked from Zakaria/hermes-agent
Code Pull Requests Activity

Hermes-agent

Browse Source
This commit is contained in:
Zakaria
2026-06-14 14:30:48 -04:00
commit dac4b88b94
5058 changed files with 1884848 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/x_search_tool.py
+525
View File
@@ -0,0 +1,525 @@
#!/usr/bin/env python3
"""X Search tool backed by xAI's built-in ``x_search`` Responses API tool.
Authentication
--------------
The tool registers when **either** xAI credential path is available:
* ``XAI_API_KEY`` is set in ``~/.hermes/.env`` or the process environment
(paid xAI API key), OR
* The user is signed in via xAI Grok OAuth — SuperGrok subscription —
i.e. ``hermes auth add xai-oauth`` has been run and the stored refresh
token still works.
Credential preference at call time matches
:func:`tools.xai_http.resolve_xai_http_credentials`: SuperGrok OAuth first,
direct OAuth resolver second, ``XAI_API_KEY`` last. That helper also
auto-refreshes the OAuth access token when it's within the refresh skew
window, so a ``True`` from :func:`check_x_search_requirements` means the
bearer is fetchable AND non-empty.
Defensive output
----------------
The tool surfaces two additional signals beyond xAI's raw response so callers
can tell a real citation-backed answer from an unsourced one:
* ``from_date`` / ``to_date`` are validated client-side before the HTTP call.
Malformed (non ``YYYY-MM-DD``), inverted (``from_date > to_date``), and
pure-future ranges (``from_date`` later than today UTC) fail fast with a
clear error instead of burning an API call. ``to_date`` in the future is
still allowed so callers can legitimately request "from yesterday to
tomorrow".
* Successful responses carry ``degraded`` and ``degraded_reason`` fields.
``degraded`` is ``True`` when any narrowing filter (handles or dates) was
active AND xAI returned no citations in either the top-level ``citations``
array or the inline ``url_citation`` annotations. In that case the
``answer`` came from the model's own knowledge rather than the X index,
and the caller should treat the result as unsourced.
Salvaged from PR #10786 (originally by @Jaaneek); credential resolution
reworked to honor both auth modes per Teknium's design.
"""
from __future__ import annotations
import json
import logging
import time
from datetime import date, datetime, timezone
from typing import Any, Dict, List, Optional, Tuple
import requests
from tools.registry import registry, tool_error
from tools.xai_http import hermes_xai_user_agent, resolve_xai_http_credentials
logger = logging.getLogger(__name__)
DEFAULT_XAI_BASE_URL = "https://api.x.ai/v1"
DEFAULT_X_SEARCH_MODEL = "grok-4.20-reasoning"
DEFAULT_X_SEARCH_TIMEOUT_SECONDS = 180
DEFAULT_X_SEARCH_RETRIES = 2
MAX_HANDLES = 10
# ---------------------------------------------------------------------------
# Config
# ---------------------------------------------------------------------------
def _load_x_search_config() -> Dict[str, Any]:
try:
from hermes_cli.config import load_config
return load_config().get("x_search", {}) or {}
except Exception:
return {}
def _get_x_search_model() -> str:
cfg = _load_x_search_config()
return (str(cfg.get("model") or "").strip() or DEFAULT_X_SEARCH_MODEL)
def _get_x_search_timeout_seconds() -> int:
cfg = _load_x_search_config()
raw_value = cfg.get("timeout_seconds", DEFAULT_X_SEARCH_TIMEOUT_SECONDS)
try:
return max(30, int(raw_value))
except Exception:
return DEFAULT_X_SEARCH_TIMEOUT_SECONDS
def _get_x_search_retries() -> int:
cfg = _load_x_search_config()
raw_value = cfg.get("retries", DEFAULT_X_SEARCH_RETRIES)
try:
return max(0, int(raw_value))
except Exception:
return DEFAULT_X_SEARCH_RETRIES
# ---------------------------------------------------------------------------
# Credential resolution
# ---------------------------------------------------------------------------
def _resolve_xai_bearer() -> Tuple[str, str, str]:
"""Return ``(api_key, base_url, source)``.
``source`` is one of ``"xai-oauth"`` or ``"xai"`` so callers (and tests)
can tell which credential path won. Raises ``RuntimeError`` if no usable
credential is available — the registered :func:`check_x_search_requirements`
gate makes that case unreachable in normal operation, but the runtime
check exists so a credential that expires between registration and
invocation produces a clean tool error instead of a 401.
"""
creds = resolve_xai_http_credentials()
api_key = str(creds.get("api_key") or "").strip()
if not api_key:
raise RuntimeError(
"No xAI credentials available. Run `hermes auth add xai-oauth` "
"to sign in with your SuperGrok subscription, or set XAI_API_KEY."
)
base_url = str(creds.get("base_url") or DEFAULT_XAI_BASE_URL).strip().rstrip("/")
source = str(creds.get("provider") or "xai")
return api_key, base_url, source
def check_x_search_requirements() -> bool:
"""Return True when xAI credentials are available AND valid.
``resolve_xai_http_credentials`` calls
:func:`hermes_cli.auth.resolve_xai_oauth_runtime_credentials` which
auto-refreshes the OAuth access token if it's expiring; a successful
return therefore implies a usable bearer.
"""
try:
creds = resolve_xai_http_credentials()
return bool(str(creds.get("api_key") or "").strip())
except Exception:
return False
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def _normalize_handles(handles: Optional[List[str]], field_name: str) -> List[str]:
cleaned: List[str] = []
for handle in handles or []:
normalized = str(handle or "").strip().lstrip("@")
if normalized:
cleaned.append(normalized)
if len(cleaned) > MAX_HANDLES:
raise ValueError(f"{field_name} supports at most {MAX_HANDLES} handles")
return cleaned
def _parse_iso_date(value: str, field_name: str) -> date:
"""Parse a strict YYYY-MM-DD string into a ``date``.
xAI accepts any string in the ``from_date``/``to_date`` slots and silently
returns an answer with no citations when the value is malformed or refers
to a window where no posts can exist. That behavior burns a billable API
call and produces a confident-sounding fluff answer that's hard for callers
to distinguish from a real result. Validating client-side fails fast and
gives the agent a clear error to act on.
"""
raw = value.strip()
try:
return datetime.strptime(raw, "%Y-%m-%d").date()
except ValueError as exc:
raise ValueError(
f"{field_name} must be YYYY-MM-DD (got {raw!r})"
) from exc
def _validate_date_range(from_date: str, to_date: str) -> None:
"""Validate ``from_date`` / ``to_date`` before they reach xAI.
Rules:
* Either field, if non-empty, must parse as ``YYYY-MM-DD``.
* When both are set, ``from_date <= to_date``.
* ``from_date`` must not be later than today UTC — no posts can exist
in a window that hasn't started yet, so the call would be guaranteed
to return zero citations. ``to_date`` in the future is allowed
(callers may legitimately set "from yesterday to tomorrow").
"""
parsed_from: Optional[date] = None
parsed_to: Optional[date] = None
if from_date.strip():
parsed_from = _parse_iso_date(from_date, "from_date")
if to_date.strip():
parsed_to = _parse_iso_date(to_date, "to_date")
if parsed_from and parsed_to and parsed_from > parsed_to:
raise ValueError(
f"from_date ({parsed_from.isoformat()}) must be on or before "
f"to_date ({parsed_to.isoformat()})"
)
if parsed_from is not None:
today_utc = datetime.now(timezone.utc).date()
if parsed_from > today_utc:
raise ValueError(
f"from_date ({parsed_from.isoformat()}) is in the future; "
f"X Search only indexes past posts (today UTC is "
f"{today_utc.isoformat()})"
)
def _extract_response_text(payload: Dict[str, Any]) -> str:
output_text = str(payload.get("output_text") or "").strip()
if output_text:
return output_text
parts: List[str] = []
for item in payload.get("output", []) or []:
if item.get("type") != "message":
continue
for content in item.get("content", []) or []:
ctype = content.get("type")
if ctype in {"output_text", "text"}:
text = str(content.get("text") or "").strip()
if text:
parts.append(text)
return "\n\n".join(parts).strip()
def _extract_inline_citations(payload: Dict[str, Any]) -> List[Dict[str, Any]]:
citations: List[Dict[str, Any]] = []
for item in payload.get("output", []) or []:
if item.get("type") != "message":
continue
for content in item.get("content", []) or []:
for annotation in content.get("annotations", []) or []:
if annotation.get("type") != "url_citation":
continue
citations.append(
{
"url": annotation.get("url", ""),
"title": annotation.get("title", ""),
"start_index": annotation.get("start_index"),
"end_index": annotation.get("end_index"),
}
)
return citations
def _http_error_message(exc: requests.HTTPError) -> str:
response = getattr(exc, "response", None)
if response is None:
return str(exc)
try:
payload = response.json()
except Exception:
payload = None
if isinstance(payload, dict):
code = str(payload.get("code") or "").strip()
error = str(payload.get("error") or "").strip()
message = error or str(payload)
if code and code not in message:
message = f"{code}: {message}"
return message or str(exc)
text = str(getattr(response, "text", "") or "").strip()
if text:
return text[:500]
return str(exc)
# ---------------------------------------------------------------------------
# Tool implementation
# ---------------------------------------------------------------------------
def x_search_tool(
query: str,
allowed_x_handles: Optional[List[str]] = None,
excluded_x_handles: Optional[List[str]] = None,
from_date: str = "",
to_date: str = "",
enable_image_understanding: bool = False,
enable_video_understanding: bool = False,
) -> str:
if not query or not query.strip():
return tool_error("query is required for x_search")
try:
api_key, base_url, source = _resolve_xai_bearer()
except RuntimeError as exc:
return tool_error(str(exc))
try:
allowed = _normalize_handles(allowed_x_handles, "allowed_x_handles")
excluded = _normalize_handles(excluded_x_handles, "excluded_x_handles")
if allowed and excluded:
return tool_error("allowed_x_handles and excluded_x_handles cannot be used together")
try:
_validate_date_range(from_date, to_date)
except ValueError as exc:
return tool_error(str(exc))
tool_def: Dict[str, Any] = {"type": "x_search"}
if allowed:
tool_def["allowed_x_handles"] = allowed
if excluded:
tool_def["excluded_x_handles"] = excluded
if from_date.strip():
tool_def["from_date"] = from_date.strip()
if to_date.strip():
tool_def["to_date"] = to_date.strip()
if enable_image_understanding:
tool_def["enable_image_understanding"] = True
if enable_video_understanding:
tool_def["enable_video_understanding"] = True
payload = {
"model": _get_x_search_model(),
"input": [
{
"role": "user",
"content": query.strip(),
}
],
"tools": [tool_def],
"store": False,
}
timeout_seconds = _get_x_search_timeout_seconds()
max_retries = _get_x_search_retries()
response: Optional[requests.Response] = None
for attempt in range(max_retries + 1):
try:
response = requests.post(
f"{base_url}/responses",
headers={
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
"User-Agent": hermes_xai_user_agent(),
},
json=payload,
timeout=timeout_seconds,
)
response.raise_for_status()
break
except requests.HTTPError as e:
status_code = getattr(getattr(e, "response", None), "status_code", None)
if status_code is None or status_code < 500 or attempt >= max_retries:
raise
logger.warning(
"x_search upstream failure on attempt %s/%s: %s",
attempt + 1,
max_retries + 1,
_http_error_message(e),
)
time.sleep(min(5.0, 1.5 * (attempt + 1)))
except (requests.ReadTimeout, requests.ConnectionError) as e:
if attempt >= max_retries:
raise
logger.warning(
"x_search transient failure on attempt %s/%s: %s",
attempt + 1,
max_retries + 1,
e,
)
time.sleep(min(5.0, 1.5 * (attempt + 1)))
if response is None:
raise RuntimeError("x_search request did not return a response")
data = response.json()
answer = _extract_response_text(data)
citations = list(data.get("citations") or [])
inline_citations = _extract_inline_citations(data)
# Degraded-result detection.
#
# xAI returns 200 OK with a synthesized answer even when its X index
# has no posts matching the caller's narrowing filters. The answer
# then comes from the model's training data, which is misleading
# because it looks identical to a real, citation-backed result. When
# any narrowing filter is active AND both citation channels came back
# empty, mark the response as degraded so callers can decide to
# broaden filters, retry, or fall back to a different source.
active_filters: List[str] = []
if allowed:
active_filters.append("allowed_x_handles")
if excluded:
active_filters.append("excluded_x_handles")
if from_date.strip():
active_filters.append("from_date")
if to_date.strip():
active_filters.append("to_date")
degraded = bool(active_filters) and not citations and not inline_citations
degraded_reason = (
f"no citations returned despite filters: {', '.join(active_filters)}"
if degraded
else None
)
return json.dumps(
{
"success": True,
"provider": "xai",
"credential_source": source,
"tool": "x_search",
"model": payload["model"],
"query": query.strip(),
"answer": answer,
"citations": citations,
"inline_citations": inline_citations,
"degraded": degraded,
"degraded_reason": degraded_reason,
},
ensure_ascii=False,
)
except requests.HTTPError as e:
logger.error("x_search failed: %s", e, exc_info=True)
return json.dumps(
{
"success": False,
"provider": "xai",
"tool": "x_search",
"error": _http_error_message(e),
"error_type": type(e).__name__,
},
ensure_ascii=False,
)
except requests.ReadTimeout as e:
logger.error("x_search timed out: %s", e, exc_info=True)
return json.dumps(
{
"success": False,
"provider": "xai",
"tool": "x_search",
"error": f"xAI x_search timed out after {_get_x_search_timeout_seconds()} seconds",
"error_type": type(e).__name__,
},
ensure_ascii=False,
)
except Exception as e:
logger.error("x_search failed: %s", e, exc_info=True)
return json.dumps(
{
"success": False,
"provider": "xai",
"tool": "x_search",
"error": str(e),
"error_type": type(e).__name__,
},
ensure_ascii=False,
)
X_SEARCH_SCHEMA = {
"name": "x_search",
"description": (
"Search X (Twitter) posts, profiles, and threads using xAI's built-in "
"X Search tool. Use this for current discussion, reactions, or claims "
"on X rather than general web pages. Available when xAI credentials "
"are configured (SuperGrok OAuth or XAI_API_KEY)."
),
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "What to look up on X.",
},
"allowed_x_handles": {
"type": "array",
"items": {"type": "string"},
"description": "Optional list of X handles to include exclusively (max 10).",
},
"excluded_x_handles": {
"type": "array",
"items": {"type": "string"},
"description": "Optional list of X handles to exclude (max 10).",
},
"from_date": {
"type": "string",
"description": "Optional start date in YYYY-MM-DD format.",
},
"to_date": {
"type": "string",
"description": "Optional end date in YYYY-MM-DD format.",
},
"enable_image_understanding": {
"type": "boolean",
"description": "Whether xAI should analyze images attached to matching X posts.",
"default": False,
},
"enable_video_understanding": {
"type": "boolean",
"description": "Whether xAI should analyze videos attached to matching X posts.",
"default": False,
},
},
"required": ["query"],
},
}
def _handle_x_search(args, **kw):
return x_search_tool(
query=args.get("query", ""),
allowed_x_handles=args.get("allowed_x_handles"),
excluded_x_handles=args.get("excluded_x_handles"),
from_date=args.get("from_date", ""),
to_date=args.get("to_date", ""),
enable_image_understanding=bool(args.get("enable_image_understanding", False)),
enable_video_understanding=bool(args.get("enable_video_understanding", False)),
)
registry.register(
name="x_search",
toolset="x_search",
schema=X_SEARCH_SCHEMA,
handler=_handle_x_search,
check_fn=check_x_search_requirements,
requires_env=["XAI_API_KEY"],
emoji="🐦",
max_result_size_chars=100_000,
)