awbrowse
A real browser, from an agent, over a service you host.
pip install awbrowse
from awbrowse import BrowseClient
b = BrowseClient("https://browser.example.com", token="...")
page = b.browse("https://example.com")
print(page.ok, page.engine, len(page.text))
awbrowse get https://example.com # render, print the text
awbrowse shot https://example.com # -> screenshot.jpg (extension sniffed)
awbrowse --self-test # prove the contract, offline
The service origin comes from --url or AWBROWSE_URL, the token from --token
or AWBROWSE_TOKEN. Neither is guessed — a client that quietly falls back to
some default endpoint is sending your pages somewhere you did not choose.
What this is, and what it is not
It is a client. The rendering engine — headless browsers, session pools, scrape schemas — stays where it is; this is the wire contract, packaged so anything can speak it. Point it at any service exposing the routes below.
That split is deliberate. The alternative was lifting a 1,200-line service with
a dozen private imports into a package, which produces something that
ModuleNotFoundErrors on your machine while reading as authoritative. A broken
package is worse than no package.
| route | what it does |
|---|---|
POST /browse |
render one page |
POST /scrape |
structured extraction |
POST /session/open |
a persistent, authenticated session |
POST /session/{id}/act |
click, type, navigate |
POST /session/{id}/observe |
read the page back |
GET /session/{id}/network |
what the page actually requested |
Authenticated pages go through a session
browse() cannot take cookies, and that is a feature.
The service's browse model declares extra="forbid". It did not always: with
pydantic's default extra="ignore", a request carrying cookies or
storage_state had those fields silently dropped, so an "authenticated"
render ran anonymous and returned the login page as though it were the app —
HTTP 200, correct-looking output, completely wrong answer.
So this client sends exactly the four declared fields and nothing else, and credentials go where the server actually reads them:
with b.open_session(storage_state=...) as s:
s.fill("#user", "me") # value= , not text=
s.click("#login")
o = s.observe() # url, title, text, ELEMENTS, screenshot
for el in o.elements: # what an agent decides to click from
print(el["tag"], el["text"], el["selector"])
for r in s.network(clear=False): # why a page that "renders fine" is failing
print(r["status"], r["content_type"], r["url"])
session.network() is the reason to reach for a session while debugging: a page
that renders correctly and fails silently is usually failing in a request nobody
looked at. Two things make an empty list mean nothing, and both are measured:
it is FILTERED (only JSON-ish responses — content-type json, /api/ in the URL,
or .json; never a 4xx/5xx and never a preflight), and the default DRAINS the
buffer server-side, so a second consecutive call returns nothing.
observe() is a different shape from browse() — it has a title and an
elements list that /browse does not send. They are separate types for that
reason: parsed as a Page, elements vanishes and an agent picks what to click
from a list it never saw.
An action the service does not dispatch is refused here, because it would not be
refused there: the server's dispatch is an if/elif chain with no else, so a
typo'd action is a no-op that answers 200 — "I clicked and nothing happened".
Same for a misspelt open_session field: unlike /browse, that model does not
forbid extras, so a mistyped storage_state is dropped in silence and hands you
an anonymous session that looks logged in.
What a live probe found that no offline test could
Both of these shipped in the first draft, and both are the failure this package exists to refuse — a request that is valid, a response that is 200, and a key that is simply absent:
The screenshot key is not the screenshot field. The request says
screenshot: true; the response says screenshot_base64. Reading the
request's name back off the response yields None on every successful capture,
so awbrowse shot would have reported "the service returned no screenshot"
forever while the service sent one every time.
The capture is JPEG, not PNG. Nothing in the request says so. A default
output of screenshot.png writes JPEG bytes under a name that lies about them —
fine in a viewer that sniffs, broken in anything that trusts the extension. So
the extension is sniffed from the magic bytes, and an -o whose extension
disagrees is honoured and mentioned.
Page therefore exposes exactly what the service was measured to send —
status, url, engine, content, and screenshot_base64 when asked. No
title, no html: this route sends neither, and an attribute that is empty on
every response is worse than an absent one, because callers branch on it.
Everything else is on .raw.
Two things it refuses to do
Return an empty page on failure. A dead service raises BrowseError. If it
returned "", a transport failure and a genuinely blank page would be the same
value to every caller, and the outage would look like a boring page.
Send an empty Authorization header. No token means no header at all. An
empty Bearer is rejected differently from an absent one, and the difference sends
you debugging the auth server instead of your config.
--self-test
Every install can prove the client still holds its contract, with no service and no network:
$ awbrowse --self-test
PASS browse body is exactly the declared field set
PASS text reads `content`; screenshot reads the MEASURED response key
PASS a failed render raises rather than arriving as a blank page
PASS network reads the MEASURED key; an undispatched action is refused
PASS image format sniffed, not assumed; base_url normalised; no empty Bearer
SELF-TEST: awbrowse ok
It asserts the exact request field set rather than membership — an extra key is a 422 for the whole request, and a test that only checks "url is in there" passes on that bug.
The aw family
Standalone tools that share one idea: replace something you would otherwise have to trust with something you can check.
Each installs on its own, works offline, and needs no account.
| instead of trusting | you check | |
|---|---|---|
| awdk | a framework's idea of how your agents should run | one loop you can read, pointed at a backend you already pay for |
| awskills | that an agent knows your procedure | the procedure written down, versioned, and loadable by any agent |
| awm | that memory stayed in its lane | tenant:user:project scopes, so a write cannot cross a boundary |
| awnode | a vendor's cloud with every prompt | a local gateway routing to backends you chose |
| awgraph | that grep found everything | an AST + tree-sitter call graph an agent can traverse |
| awgit | that no one else is editing this file | a lease, refused at commit time if you do not hold it |
| awseal | that the artifact came from who you think | an Ed25519 seal — the key that verifies is not the key that forges |
| awshare | that the download is intact | content-addressed bundles, verified on fetch |
| awnest | that there is a person on the other end | a verdict with evidence, where "we could not tell" is not "yes" |
| awnboard | a share link anyone who sees it can use | an invitation addressed to one person, for one gate, revocable |
| awnix | that the box is what you left it as | an immutable image you built, with atomic rollback |
| awrecover | that the restore worked | a restore that fully lands or does not land at all |
| awrelay | a SaaS in the middle of your agents | findings, alerts and coordination over your own transport |
| awmail | a mailbox somebody else can read | mail your agents send and receive over your own server |
| awfind | one vendor's idea of the web | results from whichever providers you configured |
| awbrowse (you are here) | that the page said what you were told | the render, the DOM and the requests it made |
| aitherkvcache | a vendor's quantisation defaults | sub-byte KV cache kernels you can benchmark yourself |
| AitherZero | a pile of scripts nobody has numbered | numbered, discoverable automation with declarative playbooks |
| AitherConnect | what a page tells your browser to do | a federated search and desktop bridge you host |
| awreason | a confident paragraph | the phases it went through, and every tool call it made to get there |
| awrecurse | that everything you pasted in was actually read | which slices it opened, and what it concluded from each |
| awprism | the first explanation that fits | the ranked alternatives, and the observation that separates them |
| awrepl | what the agent believes the value is | the value, printed from the live session |
| awresearch | a summary of pages nobody opened | every claim against the source it came from |
| awpredict | a model because it trained without erroring | its prediction against a self-updating lookup, on the rows that are actually novel |
| awkno | that the docs site is up, or that you remember the family | the whole ecosystem in your terminal, with no network at all |
awnix is the ground floor — A Linux you can hand to an agent — immutable base, capabilities included.
The Aitherium ecosystem
Every repository here is public. Each publishes an aither-manifest.json beside its page, so any surface can read every sibling's — the network is browsable from any node in it.
| repo | what it is | pages |
|---|---|---|
| awdk | Build AI agent fleets — 3 lines, any backend, local or cloud | docs |
| awskills | Portable agent skills — self-contained procedures an agent loads on demand | docs |
| awm | A portable, scoped agent memory | docs |
| awnode | A lightweight local gateway — bridges your apps to the AI backends you chose | docs |
| awrun | A priority-aware queue and dispatcher for agentic runs and ad-hoc CI builds | docs |
| awgraph | A semantic code graph for agents — AST + tree-sitter, call graphs | docs |
| awgit | Semantic version control on top of git — edit-ops and leases | docs |
| awseal | Sign an artifact so a stranger can verify it | docs |
| awshare | Publish an artifact and fetch it back verified | docs |
| awdit | An append-only audit trail whose gaps are DETECTABLE | docs |
| awbac | Role-based access control that fails closed and explains itself | docs |
| awiam | Who is this caller? A directory and session store that fails honestly | docs |
| awtunnel | Reach a service that has no public address | docs |
| awnest | Prove there is a human before you let them into the nest | docs |
| awnboard | A front gate you can put in front of anything, and hand someone the key to | docs |
| awnix | A Linux you can hand to an agent — immutable base, capabilities included | docs |
| awrecover | Labelled snapshots with an all-or-nothing restore | docs |
| awrelay | Portable agent messaging — findings, alerts, coordination | docs |
| awmail | Give an agent an email address — send, and actually receive | docs |
| awnet | The agentic web — agents host a mesh, and agents join one | docs |
| awfind | A portable search client — query, results, ranking | docs |
| awbrowse (you are here) | A portable browser client — navigate, console, network, DOM, screenshot | docs |
| awknowledge | How to run a coding agent so the result survives — the laws, with evidence | docs |
| aitherkvcache | Near-optimal KV cache quantization for LLM inference — sub-byte compression | docs |
| AitherZero | PowerShell 7+ automation framework — numbered, self-describing scripts | docs |
| AitherConnect | Browser extension — federated AI search, page context, and the Living OS overlay | docs |
| awreason | A portable reasoning client — sessions, phases, thoughts, and the chain that produced the answer | docs |
| awrecurse | Answer a question over a context far larger than the window — recursively, with the trace kept | docs |
| awprism | Turn a failure into ranked hypotheses — and say what would confirm each one | docs |
| awrepl | A REPL an agent can actually use — state that survives between turns | docs |
| awresearch | Ask a research question, get a cited report you can check | docs |
| awpredict | Predict what your environment does next, and how surprised you were | docs |
| awkno | The man page for the Aither World — every brick, stack and law, offline | docs |
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file awbrowse-0.1.0.tar.gz.
File metadata
- Download URL: awbrowse-0.1.0.tar.gz
- Upload date:
- Size: 32.3 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.12.10
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
2e5503984c794496a282b2629730d82815de0a05ed8a6445aa56b13c65a2821c
|
|
| MD5 |
408d47860a6d509d726e7bceec08d1ea
|
|
| BLAKE2b-256 |
573d01089d4b9b75d0110d3a0aa2177b9ca54778a028f1765c9712a470b5f80f
|
File details
Details for the file awbrowse-0.1.0-py3-none-any.whl.
File metadata
- Download URL: awbrowse-0.1.0-py3-none-any.whl
- Upload date:
- Size: 24.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.2.0 CPython/3.12.10
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f64ddd8c4d26f62c63553ebdcda4a828dde2dc77fe58d621d97f4ff4bc44b7f9
|
|
| MD5 |
9d5f0dee05fcea883e2b0f8461a8e921
|
|
| BLAKE2b-256 |
1b826d43735c2ebfeea5c235e4d6588da0cb4f69a895822007f4f6295667921f
|