I’ve personnally been busy elsewhere, building some cool projects with others.

Today as seen some long due fixes and maintenance

• the http://www.seeks-project.info/search.php and http://www.seeks-project.info/search_exp.php are back
• some fixes and patches on the ‘experimental’ branch.

More than ever, help is appreciated on the project. Working on other projects as well should hopefully mean ways of funding the last bits of the DHT, maybe bounties for developpers, etc…

Feedback and ideas always welcome!

(Comments are broken, apologies for this, try contacting us if you have something to say).

Current release is HIPPY: P2P collaborative filter on top of metasearch engine. Additionnally Seeks implements a websearch proxy.

Seeks runs on GNU/Linux and most of *BSD. There is no windows port yet.

• Stable HIPPY release v0.4.1 is available, and conforms to the API API for collaborative filtering and recommendation of results (04/03/2012), a rock-solid release ready for production, see ChangeLog.
  • Download sources from here.
  • Download sources from Sourceforge.
• Packages are available for Ubuntu from launchpad, Arch from aur, Debian from sileht’s repository, Fedora from WilQu’s repository.
• Experimental release is available from our git repository, see the download section.
• Nightly generated tarballs from our ‘experimental’ branch are available from here.

See the Documentation for installation and configuration.

Traffic on seeks.fr has increased to the point that we face two problems:
* we get throttled down by Google (that we use in the background);
* we do not comply with the TOS of underlying search engines such as Google that prevent third parties to re-rank or mix their search results.

Therefore, we’ve decided to improve seeks.fr because it appears to be useful to a bunch of people, and to do so we are taking the two following actions, as a first step toward better service and functionalities:
* we are studying the traffic during a short period of time (about a month) in order to understand where users are coming from, what they’re doing (are they coming once and leaving ? Are they finding what they want, and in how many steps ? Are they using seeks.fr because they use Windows ?, …);
* we are switching the server to using the Bing API (using a free key) since it is the only API out there that allows the re-ordering & mixing of textual search results.

Our aim is to offer a better Seeks.fr for common users, with two goals in mind:
* experiment new things in search, such as sharing and P2P enhanced results;
* build a strong Seeks node that will benefit the entire community (through the P2P calls issued from other nodes).

Hope it helps. Let us know if you have more questions.

1. les utilisateurs du site directement : environ 12000 requêtes par jour;
2. les utilisateurs par l’intermédiaire du pair-à-pair (P2P) en utilisant d’autres instances de Seeks, notamment privées : environ 5500 requêtes par jour.

Le site fait aussi face à quelques problèmes :

1. le nombre de requêtes et leur fréquence commence à dépasser celui autorisé par les moteurs de recherche sous-jacents, et notamment Google
2. la légalité du site est discutable car Seeks dispose des résultats des autres moteurs, de manière ‘sauvage’.

Etant donnée l’augmentation de l’utilisation du site, nous avons décidé de solutionner ces problèmes et de faire de Seeks.fr un terrain d’expérimentation ouvert au public, afin d’améliorer la réflexion sur notre utilisation des moteurs de recherche, leur amélioration, la protection des données privées et le partage de l’information.

Les deux premières étapes sont les suivantes :

1. Etudier, avec son accord, la population des utilisateurs de Seeks.fr de facon anonyme (mais en voyant passer néanmoins les adresses IPs) afin de mieux cerner leurs habitudes (recherche de texte, d’images, autre ?), leur provenance (France ou ailleurs), la fréquence des requêtes (les pics, …)
2. Remplacer les moteurs sous-jacent par des appels à l’API Bing. En effet, seule l’API de Bing donne le droit de réordonner et d’enrichir les résultats. L’API de recherche de Google a été refermée ces dernières années, est payante, et ne permet pas le réordonnement ni le mixage des résultats

La première étape est un peu paradoxale, nous souhaitons en savoir plus sur les utilisateurs de Seeks.fr, et pour cela nous allons leur demander de nous laisser le faire, ou bien de se soustraire à l’étude par un bouton pour cela. Nous avons besoin d’en savoir plus sur d’ou les utilisateurs se connectent, quel systèmes ils utilisent (par exemple, il n’y a pas de version Windows de Seeks, ce qui rend légitime l’utilisation de Seeks.fr par ces utilsateurs), s’ils se connectent réguliérement dans la journée, ou s’ils essaient tous une fois puis s’en vont. A partir de ces observations, nous construirons un site mieux adapté.

Un Seeks.fr ouvert à plus d’utilisateurs, c’est un plus pour eux, et un plus pour le réseau P2P et ceux qui l’utilisent!

Vos commentaires sont les bienvenus!

Seeks now embeds a C implementation of the readability algorithm by Alberto Garcia. We turned into a plugin for Seeks. The plugin defines:

• An API call: under the /readable resource. Usage:

  /readable?url=your_url_encoded_url
  

  Example:

  
  http://s.s/readable?url=http%3A%2F%2Fseeks-project.info%2F
  
  

  The API call is available for both the HTTP server and the proxy modes.

• Search results rendering now embed a ‘Readable’ link that allows to get a ‘readable’ version of the sample page.

The embedding of a ‘readability’ plugin has several purposes:

• Improve the comfort of sampling search results;
• Offer an API to pre-process any Web page, directly through the proxy (a regexp config file will soon be added that allows to select which pages to make ‘readable’ automatically upon navigation);
• Most notably, paves the way for more and improved machine learning and similarity analysis functionalities, that will make good use of the reduced clutter in fetched Web data.

The ‘readable’ plugin is available from the 0.4.1 experimental branch. See the installation page for more information. Be sure to check the documentation of the project.

The development ticket contains more detailed information about the plugin and its implementation.

However, Seeks would not exist if we could believe in Yacy. We are rational people, putting a fair amount of time in our research and experiments with websearch. But we do not believe that freedom in websearch lies in the freedom of re-indexing the whole Web. As a matter of fact we believe that it does not matter who crawls the content, or who serves the results. What matters we believe are the following:

• Privacy: be served anonymously;
• Freedom to control the results: edit, reject, share, rank;
• Freedom to trust who you search with, who influences your results, ….

This is the reason why Seeks implements a P2P collaborative filter, and not a P2P indexer, as Yacy does.

Our research and that of others has led us to believe that decentralized crawling and ranking is bound to fail, see J.li & al, On the feasibility of peer-to-peer indexing and search, LNCS 2735, 2003. A simple proof of this is that Yacy does not work as a search engine. It is almost impossible to get usable results out of it. Ranking appears to be almost random. There could be improvements, but these would hit the theoretical bounds expressed in the article above anyways.

Yacy implements privacy by opting out from existing search engines such as Google and Bing. While this is beautiful and exciting endeavour, we believe it is simply reinventing the wheel, with less stamina. To maintain privacy, Seeks allows to proxy you through to the search engine, so it does not see your IP address or browser header. But Seeks will let you build rings of collaborative searchers, share, recommend stuff, modify results, and more. Because we believe freedom lies in the social treatment of information, not in the automated treatment of data.

Google for instance could allow more control of its results by users. The ‘top result selection code’ that was modified part of recent algorithmic changes, could be made open source, or at least tweakable by those interested.

So, with all the respect we have for Yacy, we do not believe it will succeed. And that is why we are building something else. We are focusing on elevating the pertinence and the control of results through a mix of machine learning and active user feedback. We do not believe Seeks is the next evolution of web search by any mean. But we believe that some pieces of it may be just opening the door to something else. Otherwise we would not wake up for it in the morning

Seeks implements a Free Software collaborative search platform released under the AGPLv3 license. This platform makes it possible to build a number of applications, ranging from a simple meta-search engine to a decentralized peer-to-peer collaborative search ring.

The 0.4 version is a major release. It proposes a new RESTful API. This API allows a large range of operations. Importantly, it allows to directly post / recommend results to a community of users, and to anonymously share search results clicks as recommendations on a network.

Since October 2011, the Seeks Project is now officially supported by a startup. The Seeks startup implements a 100% Free Software business model.

Install: http://seeks-project.info/wiki/index.php/Download
Download sources: http://www.seeks-project.info/site/releases/seeks-0.4.0.tar.gz

There are basically two non-exclusive ways to use your Seeks collaborative search ring:

• The easiest way to collaborate is to simply use your Seeks instance as a regular Web search engine. This is a passive recommendation scheme.

  For this

  use the HTTP method POST on resource /search/txt/query/snippet_id 

  where you replace query with the string of your choice and snippet_id with the result id you wish to recommend. This id is generated by Seeks, you need to get a list of results first, in which every returned result snippets comes with an id you can then reuse.

• A more active way of using it is to rely on the API to curate queries and results. This is an active recommendation scheme.

  For this

  use the HTTP method POST on resource /recommendation/query?
  url=http%3A%2F%2Fwww.example.com%2F 

  where the url parameter must be URL encoded.

The main difference between the two is that the first one only let you passively recommend the results that appear into the list of results for each of your queries, whereas the second method actively attaches results to one or more queries.. So for instance, if you are using Seeks with Google as a background generic search engine, you will get results from Google. Those results you select and click will be recommended to other users performing similar queries throughout your collaborative search ring.

So this is useful but sometimes falls short because the background search engine, even Google, falls short of your expectations. Simple example, you would like to know the five finalists to some startup competition. Very much likely you will get results about the competition itself, and from there you would locate the list of finalists with high probability.
Now, what you can do with Seeks is to add the websites of every of the five finalist to the query itself. This is what the API is for.

Using curl, to you would add to the list of best basket players of all time by doing something like

curl  -X POST 
"http://www.seeks-project.info/search_exp.php/recommendation
/five+best+basket+players+of+all+time?
url=http%3A%2F%2Fen.wikipedia.org%2Fwiki%2FMichael_Jordan&lang=en"

That’s it, we’ve just added Michael Jordan’s wikipedia page to the query ‘five best basket players of all time’.

You can check that it’s there:

curl  -X POST 
"http://www.seeks-project.info/search_exp.php/recommendation
/txt/five+best+basket+players+of+all+time?output=json"

and you would get something like:

{"query":"five best basket players of all time","npeers":2,
"qtime":0.16,"snippets":[{"id":1391458535,
"title":"Michael Jordan - Wikipedia, the free encyclopedia",
"url":"http://en.wikipedia.org/wiki/michael_jordan","summary":"","seeks_meta":1,
"seeks_score":0.239827,"rank":0,
"cite":"http://en.wikipedia.org/wiki/michael_jordan",
"engines":["seeks"],"type":"webpage","personalized":"yes","snpeers":1,"hits":1}]}

(Note that you can use the Seeks CLI to perform these operations).

Now, mixing up the passive and the active recommendation schemes you can recommend search results throughout your search ring, and get recommended new results as well.

Finally, Seeks gives you the ability to reject recommendations. But while recommendations are global, that is they are shared with all Seeks instances throughout the ring, rejection is local, only affects you. This is very much like your mailbox spam filter.

There are two calls to reject results:

• using

  HTTP method DELETE on the /search/txt/query/snippet_id 

  resource does cancel any recommendation ‘booster’ on that result snippet;

• using

  HTTP method DELETE on the /recommendation/query/snippet_id 

  removes the result snippet from database definitely.

Why two calls ? Because the first one cancels the passive recommendation scheme, whereas the second one cancels the active recommendation scheme.

With two recommendation calls, and two rejection calls for control, you’re ready to collaborate within a search ring.

If you cannot make your Seeks instance available to the public, you may be interested in learning more about cross-posting your results.

Now, sharing your results requires that your machine running Seeks be available as a server to others on the network. Most often this is not the case. If you are running Seeks on your laptop for instance, it is unlikely you do let others connect to it from the outside.

For this reason Seeks allows the cross-posting of your recommendations to a public Seeks server. Cross-posting means that every result you click or recommend is remembered on your machine, but also on the remote Seeks server of you choice. The remote Seeks server is available to others to connect to (by including the server into the ring). Therefore other users can benefit from your search results curation even if you are not running a server yourself.

Cross-posting is not enabled by default but we encourage you to enable it. And doing so is easy!

Locate your query-capture-config file (from the source package, in src/plugins/query_capture), and find the following section

# Seeks node URL to which to cross-post recommended results.
# default: none
cross-post-url http://www.seeks-project.info/search_exp.php

Uncomment the

cross-post-url http://www.seeks-project.info/search_exp.php

option and use the default Seeks public server, or the one of your choice.

Restart your Seeks instance, and this is it, you are now sharing your search result selection with the world!

As a last word, note that the users who will benefit from your sharing have no possible way to know that you are behind the recommended result. This is because the public Seeks server lies in between your machine and that of other users.